Visualizing Complex Data Using R

A pocket guide for ideas to unleash

your data, create killer visuals and
transform your next presentation from
good to great

N.D. Lewis
Copyright © 2013 N.D.Lewis

All rights reserved.

ISBN-13: 978-1490433547

ISBN-10: 1490433546
DEDICATION

For my lovely Angela.

OTHER BOOKS YOU WILL ALSO ENJOY

100 Statistical Tests in R: 100 Statistical Tests in R is designed to give you

rapid access to one hundred of the most popular statistical tests. It shows you,
step by step, how to carry out these tests in the free and popular R statistical
package. The book was created for the applied researcher whose primary focus
is on their subject matter rather than mathematical lemmas or statistical theory.
Step by step examples of each test are clearly described, and can be typed
directly into R as printed on the page. To accelerate your research ideas, over
three hundred applications of statistical tests across engineering, science, and
the social sciences are discussed.

For further details visit www.auscov.com

TABLE OF CONTENTS
PART 1: HOW TO IGNITE THE VISUAL POWER OF YOUR DATA
1. Unleash the visual power of your data
2. Easily illuminate associations in cross tabulations

3. Explore similarity using trees

4. How to illuminate significance in hierarchical data

5. A powerful way to simplify bivariate complexity

6. Network your way to better correlation visualization

7. How to use a Rose for directional impact

8. Maximize trends in multivariate timeseries data
PART 2: FIFTY VISULIZATIONS TO UNLEASH YOUR CREATIVITY
9. Tips on Color and Shapes in R
10. Bubble plot

11. Multivariate shaded time series plot

12. Multivariate marginal shaded time series plot

13. Multivariate split marginal shaded time series plot

14. Word cloud
15. Commonality word cloud

16. Comparison word cloud

17. State and cities with capital map
18. Geographic pie chart map
19. US state income tax map

20. US Presidential election map

21. Population Cartogram
22. Visibility map
23. Geographic map with shape observations
24. Two variable 3D scatter plot

25. 3D scatter plot

26. 3D cluster scatter plot
27. Vertical line with points 3D scatter plot
28. Vertical labeled 3D scatter plot
29. 3D grid color bar plot
30. Circle plot

31. Circular arrows split plot

32. Rose diagram

33. Stacked rose diagram

34. Radial rose diagram

35. Stacked multivariable rose diagram

36. Pollution rose

37. Polar frequency plot

38. Australian wind rose

39. Basic heat map

40. Circular data heat map with dendrograms

41. Column trace dendrogram heat map

42. Annotated heat map

43. Correlation matrix
44. Hinton plot

45. Battleship plot

46. Density rose split plot
47. Cylindrical graded bar plot

48. Centipede plot

49. Circular probability split plot
50. Wind - pollution stacked bar time series chart
51. Environmental style time series summary plot

52. Bumps chart

53. Two ordinate time series stack plot
54. Two ordinate time series plot
55. Chernoff faces
56. Textual analysis Chernoff face plot

57. Chernoff face time series plot

58. Chernoff face rank order plot
59. Constant feature Chernoff faces
 PART 1: HOW TO IGNITE THE
VISUAL POWER OF YOUR DATA

“The secret to becoming a winner is knowing

that you can be whatever you were created to
be” – Lewis Timberlake
 1. Unleash the visual power of your data
Wish you had fresh ways to present data, explore relationships, visualize high
volume datasets and break free from mundane charts and diagrams? Visualizing
complex relationships with ease begins here. In this book you will find
innovative ideas to unlock the relationships in your own data and create killer
visuals to help you transform your next presentation from good to great. This
book is for you if you:
Need to impress your boss with you next presentation!
Need fresh ideas on the presentation of complex data.
Want to boost your visual presentation toolkit.
Wish to capture and connect relationships in your data.
Uncover and exploit new insights.
Master new techniques for data presentation.
Seek real life applications of scientific data visualization.
Whittle down complexity to its bare bones.
Stimulate innovation, creativity and insight.
Intensify knowledge discovery using visual techniques.
Diagnose, segment, connect or clarify obtuse relationships.
Need inspiration to ignite your own creativity.
In every chapter, you will discover easy to use, scientifically based tools
guaranteed to help you visualize relationships in your data. The pages you are
about to read contain the key to unleash the power of your data visually. In this
book you will find straight forward practical advice. By applying the insights
from each chapter to your own data, you will discover how to create scientific
data visualizations easily and quickly, using the free and widely available R
computing package. You do not need to have any prior experience using R;
Everything is explained clearly in the text.
Altogether you will find in these pages hundreds of ideas, tips and suggestions
you can use to boost both your understanding and presentation of complex
relationships. If you make use of even 10 percent of these, you will leap light-
years beyond others in your ability to explore, comprehend and present the
relationships in your own complex data. To assist you in recreating the
illustrations used in the text the R computing packaged is used throughout.
Illustrations are fully documented with detailed instructions on how to recreate
them given at the end of each chapter. We use R, because it is free, easy to
learn, extremely popular and runs on a variety of computing platforms. You do
not need any prior knowledge of R to use this book. I have found individuals
learn R best by ‘doing’, and this book was designed to be used actively. To get
started, type in the examples as they appear in the text. You will be amazed at
how easy R is to learn by simply ‘doing’!
Scientific Data Visualization
When I read my Master’s Degree in computer visualization and PhD in Statistics
almost a decade ago, job positions in the field of scientific data visualization
were rare, existing mainly in university research departments. This is no longer
the case. A new and exciting market for scientific visualization has emerged in
recent years. Commercial companies, government agencies and research
institutions have all recognized the value in using scientific visualization tools
to explore, understand and exploit complex relationships in their data. A casual
search of a popular online jobsite using the term “Data visualization” returned
over 2,000 active open positions, with many offering basic salaries in excess of
$150,000 per year!
All of a sudden hundreds of companies want you to help them visually analyze
their data, create new visualizations, and visually display interesting
relationships of all kinds. Typically, these positions ask for the ability to:
• Develop data visualizations for exploratory and expository data
analysis.
• Create visualizations for public consumption that inform, engage, and
educate.
• Research new ways of displaying scientific data.
• Explore data structures and features to discern representation
challenges.
Good scientific visualization captures the audience’s attention and exposes
complex associations which might be easily overlooked by “traditional” non-
visual presentational tools. When used effectively scientific visualization has
the capacity to achieve two valuable goals. First, it can expose something new
about the underlying patterns and relationships contained within the data.
Second, it provides a mechanism to obtain a “big picture” perspective, by
quickly illustrating exactly how data items are interrelated.
By taking data and turning it into visual content to expose relationships, business
leaders, consultants and consumers are more likely to engage with, share and
productively exploit it. Understanding relationships and being able to observe
them is critical for good decision making. The key in communicating scientific
information lies in presenting relationships in an accessible and digestible way.
Have no doubt, if you possess the tools to do this you will stand out from the
rest of the pack.
However, not all data visualization techniques are created equal. Just as we
have paints, pencils, chalk and film to help us capture the world in different
mediums, there are multiple ways in which to investigate the same data. In
Visualizing Complex Data using R we have sorted through hundreds of
scientific visualization ideas, and present to you in these pages those which we
have found both practicable and informative, and can be implemented by you
easily and quickly using R.
The Exceptional Power of Data Visualization
To illustrate the raw power of data visualization, let’s walk through a simple
example. Imagine the following situation. You have been asked by your boss to
analyze the relationships in a large dataset, perhaps containing thousands or
even millions of rows of data. A key customer, product or profit hangs in the
balance. Your boss has to make a decision and wants to be informed by the data
before making it. You must present your result tomorrow! How are you going to
make sense of all that information efficiently?
To make things a little easier suppose you have been asked to examine the four
datasets presented in Table 1 1[i]. Your boss wants you to identify significant
differences between the datasets. What are you to do?
Table 1 1 The four dataset's you need to investigate

Recalling your undergraduate statistics 101 class, you calculate summary

statistics. The results are shown in Table 1 2 . There appears to be little
difference between the four datasets. However, just to be on the safe side
(which is the best place to be with your stress ball of a boss!), you also
calculate the correlation coefficient between each x and y variable. The results
are shown in Table 1 3 . Again, you see no visible difference between the
datasets.

Table 1 2 Summary statistics (mean, standard deviation)

Table 1 3 The x,y correlation coefficients for each dataset

If you were to end your analysis with this conclusion you would be making a
huge mistake. At best the mistake will make your boss look foolish, cost you
your job, your company lost profits and result in a very dissatisfied client. Too
see why, take a look at Figure 1 1 , Figure 1 2 , Figure 1 3 and Figure 1 4 . All
four diagrams show the scatterplot of x and y (with regression line in red) for
each dataset. Clearly the relationship between x and y differs greatly, yet this
difference was unobservable in the traditional summary statistics of mean,
standard deviation and correlation. If you were to report the results of standard
statistical analysis, you would not find much difference between the data. The
real magic happens when these datasets are represented visually. All of a
sudden, the datasets look very different from each other, and you cannot help but
see and understand much more about the underlying relationships. This is the
power of visualizing data. Scientific data visualization is a collection
statistical methods, both quantitative and qualitative, for identifying
relationships in data and consolidating them into an illustrative informative
summarizing graphic.
Figure 1 1 Plot of x and y for dataset 1

Figure 1 2 Plot of x and y for dataset 2

Figure 1 3 Plot of x and y for dataset 3
Figure 1 4 Plot of x and y for dataset 4
Totally new to R?
New users to R can use this book easily and without any prior knowledge. This
is best achieved by typing in the examples as they are given, reading the
comments and experimenting beginning with the visualization tips.
Since R is a software system which gains it power from user developed
packages, the entire code, including installation of packages are given for each
chart. Installation commands need only be entered if the package is not already
installed on your machine. If an R package mentioned in the text is not installed
on your machine you can download it by typing:
install.packages("package_name")
Followed by:
require(package_name)
For example to download and install the portfolio package you would type in
the R console window:
install.packages("portfolio")
require(portfolio)

Tip: Copies of R and free tutorial guides for beginners can be downloaded at
http://www.r-project.org/
How to get the most from this book
There are at least three ways to use this book to boost your productivity and
creativity. First, you can dip into it as an efficient reference tool. Identify the
data visualization idea you need and see how to create it quickly. For best
results type in the examples given in the text, examine the results, and then adjust
for your own use. Second, browse through the numerous illustrations, tips and
examples to help stimulate your own creativity. Third, by typing the many
illustrations and experimenting, you will strengthen your knowledge and
understanding of both scientific data visualization and R.
You don’t have to wait until you have read the entire book to incorporate the
ideas into your own analysis and presentations. You can experience their
marvelous potency for yourself almost immediately. You can go straight to the
idea of interest and immediately test, create and exploit it in your own research,
analysis and presentations.
Applying the ideas in this book will transform your analysis. If you utilize even
one tip or idea from each chapter, you will be far better prepared to win when
faced with the challenges and opportunities of the ever expanding deluge of
exploitable data. Enough talking, let’s get started!
 2. Easily illuminate associations in cross
tabulations

Figure 2 1 Marriage selection in respect to stature

Extracting meaningful relationships from cross tabulation data (also known as

contingency tables) is an important task for many data analysts. Visual
techniques offer powerful means to illuminate associations in cross tabulations.
As an illustration, consider the famous investigation on marriage selection in
respect to stature conducted by Sir Francis Galton[ii]. Data was collected on
205 married couples, divided into three height groups – Tall, Medium and Short
respectively (medium male measurements being taken as 67 inches to 70
inches). The number of marriages in each possible combination between are
presented in the cross tabulation in Figure 2 1 . On rather close inspection you
might observe that men and women of contrasted heights, Short and Tall, or Tall
and Short, married about as frequently as men and women of similar heights,
both Tall or both Short. Of course, when presenting to an eager audience, the
effort required to obtain and extract this interpretation from the cross tabulation
table may be too great. You need more efficient techniques in your toolkit to
display and present important cross tabulation associations. One solution to
easilyilluminate associations in cross tabulations is to deploy the association
plot illustrated in Figure 2 2 .
Figure 2 2 Association plot of marriage selection in respect to stature

In this plot, the area of the boxes are proportional to the difference in observed
and expected frequencies in the data. The green rectangles above the dashed
lines indicate observed frequencies exceeding those expected. The red
rectangles below the dashed lines indicate observed frequencies below those
expected[iii]. In this example, the green boxes indicate positive association, and
the red boxes indicate negative association. The size and shapes of the box
indicate how strong the associations are. It is immediately evident that tall
people tend to marry tall people at a higher rate than one might expect;
However, this is not the case with short people, they tend to marry other short
people at lower frequency than might be expected.
Figure 2 2 was created using the assocplot function contained in the R base
package. It was built using the following three steps:
Step 1: Download and install the required packages.
require{graphics}

Step 2: Clean and prepare the data

M <- as.table(cbind(c(18,28,14), c(20,51,28) , c(12,25,9)))
dimnames(M) <- list(Husband = c("Tall", "Medium", "Short"), Wife= c("Tall","Medium", "Short"))

Step 3: Plot the chart

assocplot(M,col = c("green", "red"))
Step 1: Installs the required packaged and makes it available to R.
Step 2: Creates a table comparableto that shown in Figure 2 1 .
Step 3: The assocplot function draws the image.
Tip 1: Change the color of the boxes to suite your taste. In Figure 2 2 we
used col = c("green", "red") to create green boxes for positive association and red
boxes for negative association.
Tip 2: The R package provides multiple ways to achieve a task.It is always
useful to have knowledge of an alternative function for the same task. An
association plot can be created using the assoc function contained in
the vcd package. The result is shown in Figure 2 3 .

Figure 2 3 Association plot using the assoc function in the vcd package

Figure 2 3 was constructed using the following lines of R code:

install.packages(vcd)
library(vcd)
M <- as.table(cbind(c(18,28,14), c(20,51,28) , c(12,25,9)))
dimnames(M) <- list(Husband = c("Tall", "Medium", "Short"), Wife= c("Tall","Medium", "Short"))
assoc(M)
Spline Plots
Another visualizationthat is useful for cross tabulations is the spline plot. It
shows differing sized boxes associated with the frequency of each factor in a
cross tabulation. Figure 2 4 illustrates a spline plot for marriage selection in
respect to stature. Since there are three factors (Tall, Medium and Short) the
resultant image is composed of nine shaded boxes, sized by frequency.
Figure 2 5 reports the small-pox attack rate (percentage) in
Smallpox in England:
infected homes of individuals under and over 10 years of age for five English
towns. The attack rate for those under ten appears consistently smaller across
the towns. A similar observations appears to hold for those vaccinated over ten.
These initial observations are more easily observed in the spine plot of Figure 2 6 .

Figure 24 Spline plot of marriage selection in respectto stature

Figure 2 5 Smallpox in infected homes in five English towns

Figure 2 6 was created using the spineplot function contained in the R base
package. It was built using the following three steps:
Step 1: Download and install the required packages.
require{graphics}

Step 2: Clean and prepare the data

r1 = c (7.9, 67.6, 28.3, 53.6)
r2 = c (4.4, 54.5, 29.9, 57.6)
r3 = c (10.2, 50, 27.7, 53.4)
r4 = c (2.5, 35.3, 22.2, 47)
r5 = c (8.5, 46.3, 32.2, 50)
data <- as.table(rbind(r1,r2,r3,r4,r5))

dimnames(data) <- list(Town = c("Sheffield","Warrington", "Dewsbury","Leicester","Gloucester")

, Vaccinated = c("Yes (< 10)","No (<10)", "Yes (> 10)","No (> 10)"))

Step 3: Plot the chart

spineplot(data)

Figure 2 6 Spine plot of smallpox in infected homes in five English towns

Step 1: Installs the required packaged and makes it available to R.

Step 2: Creates a table comparableto that shown in Figure 2 5 .
Step 3: The spineplot function draws the image.
Tip : You can easilychange the color of the individual boxes in the spine plot.
For example in Figure 27 we used spineplot(data,col=
c("darkred","orange","darkgreen","yellow1")) to create dark red, orange, dark green and
yellow boxes for vaccinated under 10, not vaccinated under 10, vaccinated over
ten and not vaccinated over ten respectively.

Figure 2 7 Colorful spine plot of smallpox in five English towns

Mosiac Plots
Mosiac plots offer another visualization technique for cross tabulation data.
They display relative frequencies in terms of differing sized rectangles. The
more frequent a category is observed, the larger is the corresponding
rectangle. Figure 2 8 shows the mosaic plot of small-pox data. The mosaic plot
shows many things clearly, for instance, that the attack rate is much lower for
those vaccinated below the age of ten, and that this rate is relatively lower
across all five English towns.

Figure 2 8 Mosiac plot of smallpox in five English towns

Figure 2 8 was created using the mosaicplot function contained in the R base
package. It was built using the following three steps:
Step 1: Download and install the required packages.
require{graphics}

Step 2: Clean and prepare the data

r1 = c (7.9, 67.6, 28.3, 53.6)
r2 = c (4.4, 54.5, 29.9, 57.6)
r3 = c (10.2, 50, 27.7, 53.4)
r4 = c (2.5, 35.3, 22.2, 47)
r5 = c (8.5, 46.3, 32.2, 50)
data <- as.table(rbind(r1,r2,r3,r4,r5))

dimnames(data) <- list(Town = c("Sheffield","Warrington", "Dewsbury","Leicester","Gloucester")

, Vaccinated = c("Yes (< 10)","No (<10)", "Yes (> 10)","No (> 10)"))
Step 3: Plot the chart
mosaicplot(data, color = 1:4)

Step 1: Installs the required packaged and makes it available to R.

Step 2: Creates a table comparableto that shown in Figure 2 5 .
Step 3: The mosaicplot function draws the image. The colors are specified by the
argument color = 1:4
Tip 1: Experiment with the color of the individual boxes by setting color = 2:5 .
What do you observe?

Tip 2: Occasional you may need to change the orientation of the text in the
image. This can be achieved by using the las argument. For
example, mosaicplot(data, color = 1:4, las = 1) produces the image in Figure 2 9 .
Tip 3: It is always useful to know of differentfunctions which produce similar
types of chart. A mosaic plot can also be produced using the mosaic function in
the package vcd.
Figure 2 9 Mosiac plot using mosaicplot(data, color = 1:4, las = 1)

Quick reference R code

Figure 5 1 Box plot of the effect of Vitamin C on tooth growth
data(ToothGrowth)
boxplot(len ~ dose, data = ToothGrowth,
boxwex = 0.25, at = 1:3 - 0.2,
subset = supp == "VC", col = "red",
xlab = "Vitamin C dose mg",
ylab = "Tooth length",
xlim = c(0.5, 3.5), ylim = c(0, 35), yaxs = "i")
boxplot(len ~ dose, data = ToothGrowth, add = TRUE,
boxwex = 0.25, at = 1:3 + 0.2,
subset = supp == "OJ", col = "cyan")
legend(2, 9, c("Ascorbic acid (VC)", "Orange Juice Group (OJ)"),
fill = c("red", "cyan"))

Figure 5 2 2d contour plot of the effect of Vitamin C on tooth growth

require(reshape)
require(extracat)
data(ToothGrowth)
boxplot2g(ToothGrowth$dose, ToothGrowth$len, ToothGrowth$sup)

Figure 5 3 2d contour plot of Iris data

install.packages("extracat ")
require(extracat)
attach(iris)
boxplot2g(iris$Sepal.Length,iris$Sepal.Width,iris$Species)

Figure 5 4 relationship between tips and total bill given group size
install.packages("extracat ")
install.packages("reshape")
require(extracat)
require(reshape)
data(tips)
boxplot2g(tips$total_bill,tips$tip,tips$size)

Figure 5 5 relationship between tips and total bill given day of week
install.packages("extracat ")
install.packages("reshape")
require(extracat)
require(reshape)
data(tips)
boxplot2g(tips$total_bill,tips$tip,tips$day)
Further resources
Further reading on the extracat package can be found in:
Alexander Pilhoefer, Antony Unwin (2013). New Approaches in
Visualization of Categorical Data: R Package extracat. Journal of
Statistical Software, 53(7), 1-25. URL
http://www.jstatsoft.org/v53/i07/.
For further details on the datasets mentioned in this chapter see:
Anderson, Edgar (1935). The irises of the Gaspe Peninsula, Bulletin of the
American Iris Society, 59, 2–5.

R Core Team (2013). R: A language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria. URL
http://www.R-project.org/.
Bryant, P. G. and Smith, M (1995) Practical Data Analysis: Case Studies in
Business Statistics. Homewood, IL: Richard D. Irwin Publishing.
Wtite you notes below…
 3. Explore similarity using trees

Figure 3 1 Dendrogram of major crime by various US states

A dendrogram (from Greek Dendron meaning tree, and gramma meaning

drawing) is a tree diagram used to display the groups formed by hierarchical
clustering. Hierarchical clustering is a group of statistical techniques that
measure the similarity among a group of entities. These methods start with the
calculation of the distances of each entity from all the other entities in a dataset.
Groups or clusters are then formed usually by a process of agglomeration or
division. With agglomeration algorithms, entities begin in a group of one with
close entities merged until the entire dataset is in a single group. With division
algorithms, all entities begin in a single group. They are then split into two
groups, then each group is split until all entities are in a group of one. Measures
of the distance between entities are typically calculated using Euclidean
distance. Given p variables x1,…,xp measured on n entities Euclidean distance is
calculated as:

Where xik is the value of variable xk for entity i and xjk is the value of the same
variable for entity j.
Figure 3 1 shows a dendrogram obtained from an average agglomeration method
using crime data[iv] from eleven US states. A scale which measures the
dissimilarity (distance) has been placed at the bottom of the chart. States which
are close in terms of distance have branches coded in similar colors. The
distance matrix for this data is shown in the image below. The two closest states
were Connecticut and Pennsylvania. These are distance 165 apart. The next
closet pair of states is Missouri and Alaska, which are 244 apart. Then come
New York and Vermont which are 326 apart. The District of Columbia is the
furthest apart or the least similar.
Innovative use of dendrograms
One of the major uses of dendrograms is to discern any sub- sets of observations
whose members are both close to one another and/ or isolated from the rest.
This ability has proved extremely useful in fields ranging from marketing and
fake software detection to understanding genetic diversity in agricultural crops.
dictyExpres: The dictyExpres[v] is an interactive, web-based exploratory data
analytics application providing access to over 1,000 Dictyostelium gene
expression experiments from Baylor College of Medicine. The dictyExpres
application can be used to find clusters of similarly expressed genes via
dendrogram visualization. Dendrogram visualization taken directly from the
web application is shown in Figure 3 2 .

Figure 3 2 dictyExpres dendrogram visualization

Marketing science key words: Mela et al (2013) identy trends in marketing

science through key words used in articles published in the journal “Marketing
Science “from its inception in 1982 through 2011 – 30 volumes in total. For
each journal article, the collected observations included the authors, volume,
issue and date of publication, title, and the sets of key words chosen by the
article’s author(s). Analysis included 1085 articles by 1051 authors (or co-
authors) resulting in a total of 4749 key word choices, of which 2415 were
unique. Hierarchical cluster analysis of the twenty most frequently used words
resulted in a dendrogram with four main clusters . Cluster 1 was the word
Game theory ; Cluster 2 the words Channel, Competition, Competitive,
strategy,Pricing and Retailing; Cluster 3 included the words Brand choice,
Promotion,Buyer behavior, Econometric models and Choice models; the final
cluster included the phrases Customer satisfaction, Price discrimination,
Advertising, Bayesian analysis, Hierarchical Bayes, Conjoint analysis,
Forecasting, New products and Diffusion. The researchers conclude individual
key words were not independent.
Detecting fake software: Fake antivirus software and ransom ware trick
internet users into yielding their credit card details or personal banking
information. Credit cards are then charged or banking accounts raided.
Dietrich et al (2013) use dendrograms to successfully investigate fake antivirus
software and ransom ware. Using around 213,671 screenshots of executed
malware samples, the investigators find the attacks can be can be grouped into
subsets of structurally similar images, reflecting each ransom ware family or
attack campaign.
Genetic diversity in Sudanese wheat: Fadoul et al (2013) investigate the
genetic diversity among locally cultivated wheat in Sudan. Diversity was
estimated by a Random Amplified Polymorphic DNA (RAPD) technique using
21 arbitrary 10-mer and six Dehydration Responsive Elements (DRE) primers.
The genetic distance matrix ranged from 0.435 to 0.81 and was used to construct
a dendrogram using average linkage between groups. The researchers observed
wheat cultivars were clustered into six groups; Cluster A included cultivars
Debaira, Nisr and Sasraib, cluster B included Nabta, Wadi Elniel, Argeen and
Condor, cluster C included the cultivar Imam, cluster D contained the cultivar
Al-neelain, cluster E included Tagana and Bohain and the most distant cultivar
(Khalifa) in the dendrogram stands in cluster F. The highest similarity among the
wheat cultivars was observed between Debaira and Nisr while the lowest was
between Alneelain and Bohain.
Stock market similarity: The Dow Jones Industrial Average lists 30 stocks
representative of the American market, the NASDAQ-100 is an index that tracks
the 100 largest non-financial companies in the National Association of
Securities Dealers Automated Quotations (NASDAQ) market, the FTSE100 is
an index of the 100 companies with the largest capitalization traded in the
London market, the Deutscher Aktien index (DAX) includes 30 German
companies traded in the Frankfurt market, and the SSE-50 lists the 50 major
Chinese companies traded in Shanghai. Tapinos and Mendes (2013) investigate
the similarity of these markets using dendrogram visualization. They report the
SSE-50 as the most dissimilar of all the indices.
Creating Dendrograms in R
Figure 3 1 was created using the ape package. It was constructed using data on
major crimes (Murder, Rape, Robbery, Assault, Burglary
Larceny, and Motor Theft) recorded for ten US states (Alaska,
Connecticut, Illinois, Missouri, New Mexico, New York,
Pennsylvania, Texas, Vermont and Washington) and District of
Columbia). The data is a subset of a wider sample recorded for all states by the
US census department. The dendrogram was built using the following three
steps:
Step 1: Download and install the required packages.
Install.packages(“ape”)
require(ape)

Step 2: Clean and prepare the data

raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)

Step 3: Plot the chart

plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2)
title(“Closeness of Crime across eleven US states”,line=-29,cex.main=1.8)
title(“ (US Census Department – 2005) “,line=-30,cex.main=0.8,col=”grey”)
axisPhylo(1, las = 1)

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Loads the required data into the raw.data dataframe. It contains
information on the state name and number of crimes. The data is contained in a
csv file on this books website www.auscov.com. To view raw.data in a spreadsheet
at the R prompt type:
edit(dow.jan.2005)

The image below will appear in the R consul:

The function hclust performs hierarchical cluster analysis on raw.data using a set
of distances calculated by dist(raw.data) using method “average”. The result is
converted into a tree object of class “phylo” and stored in data . Note class
“phylo” is required for printing dendrograms using ape package functionality.
Step 3: The plot function, with a number of arguments, is used to draw the
dendrogram. The argument edge.color determines the color of the edges of the tree.
To ensure similar colors for close observations we used the rainbow palette.
The color of the dendrogram text labels is determined by tip.color. The
argument edge.width determines the width or thickness ofthe edges, the larger the
thicker the edge. Finally, font is an integer specifying the type of font for the
labels: 1 = plain text, 2 = bold, 3 = italic, 4 =bold italic.
Tip 1: There is no generic way to determine the best dendrogram. It is therefore
advisable build your cluster with the function hclust using variousmethods and
then observe the difference. We used “average” , other arguments to try
include “ward”, “single”, “complete”, “mcquitty”, “median” or “centroid”.
Easily create cladograms,fans,unrooted & radial trees
A super quick way to create a cladogram, fan style plot, unrooted tree or
radialtree is to use the type argument in the plot function. Each of the respective
plots can be obtained by setting type= “cladogram” , ”fan” , ”unrooted” or “radial” . The
following lines of code illustrate each type.
We begin with a radial tree, shown in Figure 3 3 . It was constructed by typing:
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“radial”,no.margin=TRUE,cex=0.8)

Figure 3 3 Radial tree

The fan tree shown in Figure 3 4 can be easily created by entering:

plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“fan”,no.margin=TRUE)
Figure 3 4 Fan tree

Using the same method the cladogram shown in Figure 3 5 was created by
entering:
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“cladogram”,no.margin=TRUE)

Figure 3 5 Basic cladogram

Finally, the unrooted tree shown in Figure 3 6 was created by typing:
plot(data, edge.color = “red”,tip.color=”brown”,edge.width =
2,font=1,type=”unrooted”,cex=.7,no.margin=TRUE)

Figure 3 6 Simple unrooted tree

Summary
A dendrogram is a tree diagram used to display the groups formed by
hierarchical clustering. One of the major uses of dendrograms is to permit the
viewer to perform intuitive visual cluster analysis, that is, to discern any sub-
sets of observations whose members are both close to one another and/ or
isolated from the rest. This ability has proved extremely useful in fields ranging
from marketing, fake software detection to understanding genetic diversity in
agricultural crops.Flexible dendrograms can be created rapidly, easily and
efficiently using the ape package.
Quick reference R code
Figure 3 1 Dendrogram of major crime by various US states
install.packages(“ape”)
require(ape)
raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2)

title(“Closeness of Crime across eleven US states”,line=-29,cex.main=1.8)

title(“ (US Census Department – 2005) “,line=-30,cex.main=0.8,col=”grey”)
axisPhylo(1, las = 1)

Figure 3 3 Radial tree

install.packages(“ape”)
require(ape)
raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“radial”,no.margin=TRUE,cex=0.8)

Figure 3 4 Fan tree

install.packages(“ape”)
require(ape)
raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“fan”,no.margin=TRUE)

Figure 3 5 Basic cladogram

install.packages(“ape”)
require(ape)
raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)
plot(data, edge.color = rainbow(length(data$edge)/2),tip.color=”brown”,edge.width = 2,font=2,type=
“cladogram”,no.margin=TRUE)
Figure 3 6 Simple unrooted tree
install.packages(“ape”)
require(ape)
raw.data<-read.csv(“http://www.auscov.com/uploads/5/4/2/2/5422365/crime2005.csv”,header = TRUE)
rownames(raw.data)=raw.data[,1]
hc <- hclust(dist(raw.data), “average”)
data <- as.phylo(hc)
plot(data, edge.color = “red”,tip.color=”brown”,edge.width =
2,font=1,type=”unrooted”,cex=.7,no.margin=TRUE)
Further resources
Further reading on applications of dendrograms can be found in:
Dietrich, Christian J., Christian Rossow, and Norbert Pohlmann. “Exploiting
Visual Appearance to Cluster and Detect Rogue Software.” (2013).
Fadoul, Hind E., A. Marmar, and Adil A. El Hussein. “Assessment of genetic
diversity among Sudanese wheat cultivars using RAPD markers.”(2013)
Mela, Carl F., Jason Roos, and Yiting Deng. “Invited Paper—A Keyword
History of Marketing Science.” Marketing Science 32.1 (2013): 8-18.
Tapinos, Avraam, and Pedro Mendes. “A Method for Comparing Multivariate
Time Series with Different Dimensions.” PloS one 8.2 (2013): e54201.
For further details on the ape package see:
Paradis E., Claude J. & Strimmer K. 2004. APE: analyses of phylogenetics and
evolution in R language. Bioinformatics 20: 289-290.
-Wtite you notes below…
 4. How to illuminate significance in
hierarchical data

Figure 4 1 Hierarchical (tree-structured) data for the Dow-Jones Index

Treemaps display hierarchical (tree-structured) data in a colored mosaic form

containing embedded, rectangular shapes. For example, Figure 4 1 captures the
structure of Dow Jones industrial index from sector and stock data. The main
branches of the image represent the eight sectors that make up the Dow Jones
index (Communications, Cyclicals, Energy, Financials, Industrials Materials,
Staples, Technology and Utilities), with individual stocks as leaves or end
nodes. In a treemap the branch of tree structured data is given a rectangle, which
is then tiled with smaller rectangles representing sub-branches or end leaves -
see for example Figure 4 2 . A leaf node's rectangle has an area proportional to
a specified dimension of the data. The color and size dimensions of the
rectangles allow for the identification of patterns that might be difficult to spot
using standard bar charts and graphs. A key advantage of using treemaps for
large datasets is that they can legibly display thousands of items simultaneously.
Figure 4 2 Treemap of Dow-Jones Index as at the end of January 2005

Figure 4 2 presents a treemap for the Dow Jones Industrial Average. It

represents this data as eight large rectangular regions; one each for each of eight
categories making up the Dow Jones Industrial Average. The treemap partitions
the image into a collection of box-shaped elements, forming a mosaic. Each
inner box represents an individual stock (of which there are thirty in the index)
and is sized according it’s price. For example, financials consists of four stocks
and are represented by four colored rectangular boxes, each with a shading
based on the price return for January 2005. Price returns are represented using a
red through green scale; the higher the return the greener the box.
From Figure 4 2 the hierarchy of the Dow Jones Industrial Average can be
easily seen; as can the relative number of stocks for each sector. The dominant
sectors and stocks in terms of price can be quickly identified by the size of
rectangle. As can the relative performance of each sector and individual stock.
Innovative use of treemaps
This ability to represent a hierarchy of data makes the treemap visualization
both efficient and effective. Originally designed to visualize file space on very
large computer-storage disk devices, treemaps are now used in a diverse range
of fields including human-gene research, social network analysis and finance.
Treemaps are growing rapidly in popularity as data scientists, statisticians and
computer analysts discover their power to convey complex information from
extremely large data sources. We outline below a number of creative uses for
treemaps:
Visualization of news:
Newsmap[vi] visualizes the constantly changing landscape of news. It uses a
treemap visualization algorithm to picture the very large volume of information
gathered through News aggregators, see Figure 4 3. Interactions between news
items are visualized in a dynamic and constantly changing layout. The placement
of each news item and the total area it covers depends on the number of news

sources that report it.

Figure 4 3 Newsmap’s treemap visualization of US news

Information visualization in digital Libraries: The exponential growth in

digital content in libraries requires powerful tools to allow efficient navigation
of content. Wan (21013) explore the role of treemaps in the design of user-
friendly interfaces. They develop a visual treemap system to present collections
of documents to users.
Workflow management: Oliva, Gustavo, et al (2013) use treemaps to support
the management of complex workflow repositories. They are used to depict the
change impacts of the complete workflow repository, with rectangles
representing a specific section of the workflow process. The researchers report
the use of treemaps makes it easy and quick to identify repository sections that
require additional attention.
Compounds in traditional Chinese medicines: As part of a study to understand
the structural features of natural compounds from traditional Chinese medicines,
the scaffold architectures of drug-like compounds in MACCS-II Drug Data
Report (MDDR), non-drug-like compounds in Available Chemical Directory
(ACD), and natural compounds in Traditional Chinese Medicine Compound
Database (TCMCD) were explored and compared using treemaps by Tian et al
(2013). The researchers visualize the structural diversity of the scaffolds within
the overall dataset. They observe the distributions of the level 1 scaffolds of
MDDR are sparser than those of ACD and TCMCD, indicating a higher
structural diversity of the scaffolds in MDDR. They also report, some level 1
scaffolds with the highest frequencies of MDDR are similar to those of
TCMCD, indicating that some ring substructures extracted from TCMCD may
serve as valuable substructure resource for drug discovery.
Semantic web data search: Non expert users can find it difficult to explore and
use semantic web data as the available datasets are often monolithic with files
which can often only be explored using complex semantic queries. One solution
is to design user interfaces to the data, however these are often very elaborate
and difficult to navigate. Treemaps offer a natural solution because they
illuminate what kind of entities are contained in a dataset, how they are
interrelated, what are the main properties, values and so on. Brunetti et al
(2013) exploit these properties to design treemaps to make semantic web data
more accessible to non-semantic web users.
How to create a treemap using map.market
The map.market function in the package portfolio offers a simple and often the most
convenient approach to rapid treemap construction. Figure 4 2 was created
using this function. It was built using the following three steps:
Step 1: Download and install the required packages.
install.packages("portfolio")
require(portfolio)

Step 2: Clean and prepare the data

data(dow.jan.2005)

Step 3: Plot the chart

map.market(id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret)

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Loads the dow.jan.2005 dataframe in R. It contains information on the
name, stock market symbol, price and business sector of the thirty stocks in the
Dow Jones Industrial index as at the end of January 2005. To examine the
dataframe in a spreadsheet type format type:
edit(dow.jan.2005)

The image below will appear in the R consul:

Step 3: The map.market function takes four arguments. The first is the unique
identify for each data item. In the case of the dow.jan.2005 dataframe this is the
stock identification symbol obtained by dow.jan.2005$symbol . The second argument
calculates the area of the rectangles; we used the stock price contained
in dow.jan.2005$price as the size variable. The third argument is used to calculate the
color of the rectangles; the monthly return of each stock, accessed
by dow.jan.2005$month.ret and scaled by 100 was used to color.

Economic wealth & population

The dataframe GNI2010 in the package treemap contains information on population
and gross national income for 208 countries in 2010 collected by the world
bank.
Figure 4 4 Treemap of economic wealth colored by relative population

Figure 4 4 illustrates the treemap obtained using this data. It was created using
the following three steps:
Step 1: Download and install the required packages.
install.packages("portfolio")
install.packages("treemap")

Step 2: Clean and prepare the data

require("portfolio")
require("treemap")
data(GNI2010)
z=rank(GNI2010$population)
z=z-104

Step 3: Plot the chart

map.market(id = GNI2010$iso3,
area = GNI2010$GNI+1,
group = GNI2010$continent,
color = z,
main="Gross National Income - Population")

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Loads the GNI2010 dataframe into R. It contains information on the
country name, unique world bank identifier, population and gross national
income. To examine the dataframe in a spreadsheet type format type:
edit(GNI2010)

The image below will appear in the R consul:

The variable z is used to create a scale for the color of the rectangles. It is
based on the rank population size, adjusted so it lies in the range -104 to +104.
This adjustment is made to emphasize the visual contrast in colors.
Step 3: The map.market function takes four arguments. The first is the unique
identify for each data item. In this case the unique country identifier contained
obtained from GNI2010$iso3 . The second argument calculates the area of the
rectangles; we used gross national income contained in GNI2010$GNI . The
variable z measures the relative rank population size ; it is used to calculate the
color of the rectangles. Finally, the title of the chart was created
using main="Gross National Income - Population" .
Customizing map.market
Figure 4 5 Treemap of Dow-Jones Index with stock labels

Tip 1: The most common request is for the individual labels of data elements to
be represented on the treemap. This can be achieved by adding the
argument lab=TRUE to the map.market function. For example, to add labels to Figure
4 2 type:
data(dow.jan.2005)
map.market(id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret,
lab=TRUE)

The resultant image is shown in Figure 4 5.

T i p 2: It is possible to change the colors used in the map.market treemap
visualization. This requires a little hacking of the code, although it is quite
straightforward. We will change the background from grey to white, have the
text appear in blue and use yellow-green colors for positive values and orange -
dark red for negative values. To begin enter map.market at the R prompt. You
should see something along the lines of the image shown below:
This is the R code contained in the map.market function. Copy the entire output to
a text editor, feel free to use the R text editor or any other that you wish. The
first line of code looks something like this:
function (id, area, group, color, scale = NULL, lab = c(TRUE,
FALSE), main = "Map of the Market", print = TRUE)

We will create a new function to draw our required treemap. This can be
achieved in four steps:
Step 1: Add the term map.market.color<- to the start of the code. Your edited code
should now look something like this:
map.market.color<-
function (id, area, group, color, scale = NULL, lab = c(TRUE,
FALSE), main = "Map of the Market", print = TRUE)
You have just created a new function called map.market.color.
Step 2: Look for the following lines of code (you should find them around line
165 in your text editor):
mapmarket <- gTree(name = "MAPMARKET", children = gList(rectGrob(gp = gpar(col = "dark grey",
fill = "dark grey"), name = "background"), top.tree,
map.tree))
Change the phase "dark grey" to "white" . The background color has been changed
from dark grey to white!
Step 3: Now let’s change the color of the text from white to blue. Change "white"
to "blue" in the statement (around line 124):
label = this.data$label[s], gp = gpar(col = "white")
Next, change "white" to "blue" in statement (around line 133):
name = "label", gp = gpar(col = "white")

Step 4: Finally, identify the statements (around line 44):

color.ramp.pos <- colorRamp(c("black", "green"))

color.ramp.neg <- colorRamp(c("black", "red"))

These two lines of code are responsible for the color of positive and negative
numbers. Change the lines of code so that they state:
color.ramp.pos <- colorRamp(c("yellow", "darkgreen"))
color.ramp.neg <- colorRamp(c("orange", "darkred"))

Figure 4 6 Customized map.market treemap

The edited function is complete. Save it and then copy into the R consol. Now
you are ready to execute the code. Enter the following into the R consul and you
will see the image shown in Figure 4 6:
data(dow.jan.2005)
map.market.color (id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret,
lab=TRUE)
How to create a treemap using treemap
The treemap function in the package treemap provides a useful alternative for
creating a treemap in R. We illustrate the use of this function using data from the
London datastore. This is a free to access website created by the Greater
London Authority as a portal for London’s data. It is hosted at
http://data.london.gov.uk/ with home page similar to that shown below.

Let's make a withdrawal from the store. Our goal is to create a treemap using
data on the political control, income and unemployment rate for each London
borough. To do this we will use the London boroughs dataset located at:
http://data.london.gov.uk/datastore/package/london-borough-profiles . Figure
4 7 illustrates the treemap created using this dataset. It was created using the
following three steps:
Step 1: Download and install the required packages.
install.packages("treemap")
require("treemap")

Step 2: Clean and prepare the data

raw.data<-read.csv("http://data.london.gov.uk/datafiles/demographics/london-borough-profiles.csv")
data<-cbind(raw.data[,1:2],as.numeric(as.character(raw.data[,25])),as.numeric(as.character(raw.data[,31]))
data<-data[-1,]
data<-data[-1,]
data<-data[-37,]
data<-data[-36,]
data<-data[-35,]
data<-data[-34,]
data<-data[-33,]
column.names<-c("id","borough","unemp","income","control")
colnames(data)<-column.names

Step 3: Plot the chart

treemap(data,
index=c("control", "borough"),
vSize="income",
vColor="unemp",
 type="value",
palette="RdBu",
title = "London Boroughs: Political Control,Unemployment and Income",
title.legend = "Unemployment Rate (%)"
)
Explaining each step
Step 1: Installs the required packaged and makes it available to R. The require
command is necessary at the start of each new R session.

Figure 4 7 Treemap using the treemap function

Step 2: Reads a comma-separated values file (csv)from the London Datastore

into the raw.data dataframe. The dataframe consists of 69 columns of
demographic, economic and political data for all of the London Boroughs
measured in July 2011. A new dataframe named data is created using the cbind
function. It contains information on the borough name, unique identificationcode,
unemployment rate, income and political control of each borough. We get rid of
rows which do not relate to London boroughs using data<-data[-x,], where x is the
row to be deleted. Next, new and more intuitive column names are created for
the data dataframe.
Step 3: The t r e e ma p function takes several important arguments. The
argument index indicates the hierarchical structure to be used in the treemap; the
first name mentioned is the highest aggregation level, the second name the
second-highest aggregation level, and so on. We are interested in political
control by borough, and so specify "control" first followed by "borough" .The size of
the rectangles is determined by income level using vSize="income" . Color gradation
is set by vColor="unemp" using the palette "RdBu". Note the argument type takes a
wide variety of character based values. We se t type="value" which allows the
statement vColor="unemp" to be mapped directly to the "RdBu" color palette.
Summary
The ability to represent a hierarchy of data makes the treemap visualization both
efficient and effective. Originally designed to visualize file space on very large
computer-storage disk devices, treemaps are now used in a diverse range of
fields including human-gene research, social network analysis and
finance.The map.market function in the package portfolio offers a simple and often
the most convenient approach to rapid treemap construction. The treemap function
in the package treemap provides a useful alternative for custom treemaps.
Quick reference R code
Figure 4 2 Treemap of Dow-Jones Index
install.packages("portfolio")
require(portfolio)
data(dow.jan.2005)
map.market(id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret)

Figure 4 4 Treemap of economic wealth colored by relative population

install.packages("portfolio")
install.packages("treemap")
require("portfolio")
require("treemap")
data(GNI2010)
z=rank(GNI2010$population)
z=z-104
map.market(id = GNI2010$iso3,
area = GNI2010$GNI+1,
group = GNI2010$continent,
color = z,
main="Gross National Income - Population")

Figure 4 5 Treemap of Dow-Jones Index with stock

data(dow.jan.2005)
map.market(id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret,
lab=TRUE)

Figure 4 6 Customized map.market treemap-

data(dow.jan.2005)
map.market.color (id = dow.jan.2005$symbol,
area = dow.jan.2005$price,
group = dow.jan.2005$sector,
color = 100 * dow.jan.2005$month.ret,
lab=TRUE)

Figure 4 7 Treemap using the treemap function

install.packages("treemap")
require("treemap")
raw.data<-read.csv("http://data.london.gov.uk/datafiles/demographics/london-borough-profiles.csv")
data<-cbind(raw.data[,1:2],as.numeric(as.character(raw.data[,25])),as.numeric(as.character(raw.data[,31]))
data<-data[-1,]
data<-data[-1,]
data<-data[-37,]
data<-data[-36,]
data<-data[-35,]
data<-data[-34,]
data<-data[-33,]
column.names<-c("id","borough","unemp","income","control")
colnames(data)<-column.names
treemap(data,
index=c("control", "borough"),
vSize="income",
vColor="unemp",
type="value",
palette="RdBu",
title = "London Boroughs: Political Control,Unemployment and Income",
title.legend = "Unemployment Rate (%)"
)
Further resources
Further reading on applications of treemaps can be found in:
Brunetti, Josep Maria, Roberto García, and Sören Auer. "From Overview to
Facets and Pivoting for Interactive Exploration of Semantic Web Data."
International Journal on Semantic Web and Information Systems (IJSWIS) 9.1
(2013): 1-20.
Oliva, Gustavo, et al. "A Change Impact Analysis Approach for Workflow
Repository Management." ICWS-IEEE 20th International Conference on Web
Services-2013. 2013.
Tian, Sheng, et al. "Drug-likeness analysis of traditional Chinese medicines: 2.
Characterization of scaffold architectures for drug-like compounds, non-drug-
like compounds, and natural compounds from traditional Chinese medicines."
Journal of cheminformatics 5.1 (2013): 1-14.
Wan, Gang. "Visualizations for digital libraries." Information technology and
libraries 25.2 (2013): 88-94.
For further details on the map.market function see:
Jeff Enos, David Kane, with contributions from Daniel Gerlanc and
Kyle Campbell (2012). portfolio: Analyzing equity portfolios. R package
version 0.4-5. http://CRAN.R-project.org/package=portfolio
For further details on the treemap function see:
Martijn Tennekes (2013). treemap: Treemap visualization. R package
version 2.0.1. http://CRAN.R-project.org/package=treemap
Technical details on treemap construction is contained in:
Johnson, B., & Shneiderman, B. (1991) Tree-maps: A space-filling approach to
the visualization of hierarchical information structures. In Proceedings of the
2nd International IEEE Visualization Conference, pp. 284-291, San Diego, CA.
Wtite you notes below…
 5. A powerful way to simplify bivariate
complexity

Figure 5 1 Box plot of the effect of Vitamin C on tooth growth

Box plots provide a convenient way of graphically depicting groups of

numerical data through their quartiles. Figure 5 1 depicts a typical approach to
displaying the relationship between two groups using boxplots. It shows the
relationships between the length of odontoblasts (teeth) in each of 10 guinea
pigs at each of three dose levels of Vitamin C (0.5, 1, and 2 mg) using two
delivery methods (orange juice or ascorbic acid). The relationship between the
two delivery methods, dose and tooth growth is evident. Boxplots provide a
wonderful tool for displaying this type of data and should form a part of every
visual analysts toolkit.
A supplemental technique which can enhance the visual display of box plot type
data and assist in exploratory analysis is the two dimensional box plot contours.
Two dimensional box plot contours (2d-bc plot) provide a powerful way to
visualize bivariate relationships using traditional boxplot metrics. Here is how
they work. First, the sample data is centered and standardized with projections
onto direction vectors ranging from 0 to 360 degree. Second, the traditional
boxplot characteristics (median, box, whisker) are computed for each projection
and retransformed into contours using the covariance matrix.
Figure 5 2 presents a typical 2d-bc plot using the same data of the box plots of
Figure 5 1. The dose level is captured on the x-axis, with tooth length measured
by the y-axis. There appears to be a generally positive relationship across both
treatment types (VC and OJ).
One of the major uses of a 2d-bc plot is visualize intuitively relationships
between two variables which have a number of categories. Figure 5 3 shows a
2d-bc plot for sepal length (x –axis) and sepal width (y-axis) collected by
Anderson (1935) for flowers from each of 3 species of iris (setosa, versicolor,
virginica). The relationship between sepal length and width within and between
individual flower species is clearly observable.

Figure 5 2 2d contour plot of the effect of Vitamin C on tooth growth

The 2d-bc plot offers an extremely useful exploratory data analysis tool as the
following example illustrates.
Tipping in Restaurants: Wait table staff make a considerable proportion of
their income from tips. The better the service, friendlier the waiter and tastier
the food, the higher the expected tip. Bryant & Smith (1995) report on a waiter
who recorded information about each tip he received over a period of several
months. For each of the 244 tips the waiter recorded information on:
tip in dollars,
bill in dollars,
sex of the bill payer,
whether there were smokers in the party,
day of the week,
 time of day, *
size of the party.

Figure 5 3 2d contour plot of Iris data

Of great interest to managers of restaurants and wait staff is the relationship

between the size of the bill and tip in dollars given factors such as sex, day of
the week and so on. These relationships can be explored and visualized quickly
using a 2d-bc plot, for example in Figure 5 3 . We see clearly the positive
relationship between tip (x-axis) and bill size (y-axis). The relationship
between these two variables and group size is also evident, as is the
relationship between group size and bill size/tips. Thus, the 2d-bc plot provides
a powerful way to simplify bivariate group complexity.
Figure 5 4 relationship between tips and total bill given group size

As a second illustration of this powerful visualizationtool consider the

relationship between tips and total bill given the day of the week (the waiter
only worked Thursday through Sunday). This relationship is illustrated
inFigure 5 5 . There is a clear positive relationship. It is interesting to note the
contours are widest on Sunday, what do you suspect is the cause of this?
Figure 5 5 relationship between tips and total bill given day of week
Creating 2d-bc plots in R
Figure 5 3 was created using the extracat package. It was built using the following
three steps:
Step 1: Download and install the required packages.
install.packages("extracat ")-
require(extracat)

Step 2: Clean and prepare the data

attach(iris)

Step 3: Plot the chart

boxplot2g(iris$Sepal.Length,iris$Sepal.Width,iris$Species)
Step 1: Installs the required packaged and makes it available to R. If the
package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Loads the required data into the R environment. The iris data is
contained in the datasets package with comes pre-installed in R.
Step 3: The boxplot2g function draws the image. The sepal length is plotted on the
x-axis, sepal width on the y axis. The third argument passed to the function
relates to the groups or categories contained in the data, in this case species
(setosa, versicolor, virginica).
Tip : When you have a large number of categories try adding the argument colv =
rainbow(n) to boxplot2g. This will use the rainbow pallet to draw the group contours.
For example, if you had eight categories you could add colv = rainbow(8) .
Summary
The 2d-bc plot offers a powerful way to visualize the bivariate relationships
using traditional boxplot metrics. It provides an intuitive way to visualize
relationships between two variables which have a number of categories. The
chart can be used to simplify bivariate complexity for presentation purposes or
as an data exploration tool, either way it is a- technique you need in your
visualization toolbox.
Quick reference R code
Figure 5 1 Box plot of the effect of Vitamin C on tooth growth
data(ToothGrowth)
boxplot(len ~ dose, data = ToothGrowth,
boxwex = 0.25, at = 1:3 - 0.2,
subset = supp == "VC", col = "red",
xlab = "Vitamin C dose mg",
ylab = "Tooth length",
xlim = c(0.5, 3.5), ylim = c(0, 35), yaxs = "i")
boxplot(len ~ dose, data = ToothGrowth, add = TRUE,
boxwex = 0.25, at = 1:3 + 0.2,
subset = supp == "OJ", col = "cyan")
legend(2, 9, c("Ascorbic acid (VC)", "Orange Juice Group (OJ)"),
fill = c("red", "cyan"))

Figure 5 2 2d contour plot of the effect of Vitamin C on tooth growth

require(reshape)
require(extracat)
data(ToothGrowth)
boxplot2g(ToothGrowth$dose, ToothGrowth$len, ToothGrowth$sup)

Figure 5 3 2d contour plot of Iris data

install.packages("extracat ")
require(extracat)
attach(iris)
boxplot2g(iris$Sepal.Length,iris$Sepal.Width,iris$Species)

Figure 5 4 relationship between tips and total bill given group size
install.packages("extracat ")
install.packages("reshape")
require(extracat)
require(reshape)
data(tips)
boxplot2g(tips$total_bill,tips$tip,tips$size)

Figure 5 5 relationship between tips and total bill given day of week
install.packages("extracat ")
install.packages("reshape")
require(extracat)
require(reshape)
data(tips)
boxplot2g(tips$total_bill,tips$tip,tips$day)
Further resources
Further reading on the extracat package can be found in:
Alexander Pilhoefer, Antony Unwin (2013). New Approaches in
Visualization of Categorical Data: R Package extracat. Journal of
Statistical Software, 53(7), 1-25. URL
http://www.jstatsoft.org/v53/i07/.
For further details on the datasets mentioned in this chapter see:
Anderson, Edgar (1935). The irises of the Gaspe Peninsula, Bulletin of the
American Iris Society, 59, 2–5.

R Core Team (2013). R: A language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria.
URL http://www.R-project.org/.
Bryant, P. G. and Smith, M (1995) Practical Data Analysis: Case Studies in
Business Statistics. Homewood, IL: Richard D. Irwin Publishing.
Wtite you notes below…
 6. Network your way to better correlation
visualization

Figure 6 1 Typical approach used to display correlations

Exploration and display of correlations form an integral part of the day to day
duties of the data scientist. Where many variables are concerned, the usual
approach is to display a correlation table, typically similar to that shown in
Figure 6 1 . The figure shows the correlation of the daily return of eleven
exchange traded funds using closing prices over the time period August 23 rd
2011 to September 11 th 2013. Although the correlations are presented neatly, it
is often difficult to discern the true multivariate nature of the relationships. In
practice, a researcher might be interested in correlations above a certain
absolute value. A popular, although primitive, solution is to highlight such
correlations in the above table. A much better idea, is to use a correlation
network to display visually the relationship between variables.
Figure 6 2 illustrates a basic correlation network using the exchange traded
funds data. It is constructed so that green lines represent a positive correlation
between any two variables, and red lines a negative correlation. The thickness
of the lines increases with the absolute value of the correlation.
Figure 6 2 Basic correlation networks of popular exchange traded funds

The image captures neatly the multivariate nature of the correlation structure in a
more intuitive fashion that a traditional table. Further details on the correlation
structure can be revealed by only examining correlations above a specific
absolute value. For example, suppose the researcher is interested in absolute
correlations greater than 0.5. The resultant correlation network is displayed in
Figure 6 3. Notice how the multivariate interrelationships between the variables
is immediately evident. A positive core of four exchange traded funds (acw, gnr,
vwo and ivw) is apparent, whilst igv and pcy appear uncorrelated with the
remaining exchange traded funds (at the cutoff threshold).
Figure 6 3 Correlations greater than |0.5|
Innovative use of correlation networks
Physiological networks using drugs: Grossman, Adam et al (2013) built
correlation networks using physiologic data to investigate changes associated
with pressor use in intensive care. The researchers collected 29 physiological
variables at one-minute intervals from nineteen trauma patients. They grouped
each minute of data as receiving or not receiving pressors. For each group
correlation networks of pairs of physiologic variables were built. To visualize
drug-associated changes the researchers divided the resultant networks into
three components: an unchanging correlation network, a correlation network of
connections with changing correlation sign, and a correlation network of
connections only present in one group.
Mindfulness training: Taylor, Véronique A., et al (2013) examine the effect
of mindfulness training on functional brain connectivity during a restful state.
Using data from 13 experienced meditators, 11 beginner meditators and
functional magnetic resonance imaging, pairwise correlations were computed
between default mode network seed regions’ time courses. A correlation
network diagram was used to visualize the resultant correlations. The
researchers observed relative to beginners, experienced meditators had weaker
functional connectivity between default mode network (DMN) brain regions
involved in self-referential processing and emotional appraisal. However, they
noted experienced meditators had increased connectivity between certain DMN
regions, compared to beginner meditators. The researchers conclude meditation
training leads to functional connectivity changes between core DMN regions
possibly reflecting strengthened present-moment awareness.
Chinese Prices: Gao and Zhong (2013) collect data on 195 Chinese price
indices from 2003 to 2011. The relationships between the observations are
investigated using a correlation network, see Figure 6 4 . The researchers
choose a correlation of 0.82 as their threshold. Therefore If the correlation
coefficient between two price indices is greater than 0.82 there is an edge
between them.
Figure 6 4 Correlation network of Gao and Zhong[vii]
Creating correlation networks in R
Figure 6 1 was created using the qgraph package. It was built using the following
three steps:
Step 1: Download and install the required packages.
install.packages("PerformanceAnalytics")
install.packages("tseries")
install.packages ("fImport")
install.packages ("qgraph")
library(tseries)
library(PerformanceAnalytics)
library(qgraph)

Step 2: Clean and prepare the data

days = 750
date= as.Date(c("2013-09-11"))-days

con <- url("http://quote.yahoo.com")

if(!inherits(try(open(con), silent = TRUE), "try-error"))
{
close(con)
#Note if you change any of the asset classes use search and replace
#don’t do manually as you may miss something

acwi <- get.hist.quote(instrument = "acwi", start = date ,

compression = "d", quote = "Close")

vwo<- get.hist.quote(instrument = "vwo", start = date ,

compression = "d", quote = "Close")

iwv<- get.hist.quote(instrument = "ivw", start = date ,

compression = "d", quote = "Close")

mcro<- get.hist.quote(instrument = "mcro", start = date ,

compression = "d", quote = "Close")

ief<- get.hist.quote(instrument = "ief", start = date ,

compression = "d", quote = "Close")

igov<- get.hist.quote(instrument = "igov", start = date ,

compression = "d", quote = "Close")
-
pcy<- get.hist.quote(instrument = "pcy", start = date ,
compression = "d", quote = "Close")

hdg<- get.hist.quote(instrument = "hdg", start = date ,

compression = "d", quote = "Close")

rem<- get.hist.quote(instrument = "rem", start = date ,

compression = "d", quote = "Close")
fty<- get.hist.quote(instrument = "fty", start = date ,
compression = "d", quote = "Close")

gnr<- get.hist.quote(instrument = "gnr", start = date ,

compression = "d", quote = "Close")

x <- merge(acwi,vwo,iwv,mcro,ief,igov,pcy,hdg,rem,fty,gnr)

x<- na.omit(x)
colnames(x)<- c("acwi","vwo", "ivw", "mcro", "ief", "igov", "pcy", "hdg", "rem", "fty", "gnr" )

returns =diff(x)/x[-length(x)] # Calculating rates of return from price levels

returns = na.omit(returns)

# calculate correlation matrix

cor=cor(returns)

Step 3: Plot the chart

qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout= "spring",vsize=10)

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Capture the closing price of each exchange traded fund from Yahoo.
The data are converted into daily changes and cleaned, followed by the
calculation of the correlation matrix.
Step 3: The qgraph function draws the correlation network. It is constructed so
that positive correlations are displayed in dark green and negative correlations
in dark red.
Tip 1: To add a correlation cutoff value include the comment minimum=cutoff to
the qgraph function. For example, Figure 6 3 was built
using: qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout=
"spring",vsize=10,minimum=0.5)

Tip 2: The argument vsize changes node size. Try experimenting with various
values, for example set it to 20 and then 5. What do you observe?

Tip 3: A circular correlation plot can be obtained by settingthe argument layout=

"circular". For example, Figure 6 5 , was constructed using:
qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout= "circular",vsize=10)

Figure 6 5 Circular correlation network

Quick reference R code
Figure 6 2
install.packages("PerformanceAnalytics")
install.packages("tseries")
install.packages ("fImport")
install.packages -("qgraph")
library(tseries)
library(PerformanceAnalytics)
library(qgraph)

days = 750 # number of days to use in analysis approx. 3 years

date= as.Date(c("2013-09-11"))-days

library(tseries)
library(PerformanceAnalytics)

library(qgraph)

####################GET DATA FROM

YAHOO#####################################

con <- url("http://quote.yahoo.com")

if(!inherits(try(open(con), silent = TRUE), "try-error"))
{
close(con)
#Note if you change any of the asset classes use search and replace
#don’t do manually as you may miss something

acwi <- get.hist.quote(instrument = "acwi", start = date ,

compression = "d", quote = "Close")

vwo<- get.hist.quote(instrument = "vwo", start = date ,

compression = "d", quote = "Close")

iwv<- get.hist.quote(instrument = "ivw", start = date ,

compression = "d", quote = "Close")

mcro<- get.hist.quote(instrument = "mcro", start = date ,

compression = "d", quote = "Close")

ief<- get.hist.quote(instrument = "ief", start = date ,

compression = "d", quote = "Close")

igov<- get.hist.quote(instrument = "igov", start = date ,

compression = "d", quote = "Close")
-
pcy<- get.hist.quote(instrument = "pcy", start = date ,
compression = "d", quote = "Close")

hdg<- get.hist.quote(instrument = "hdg", start = date ,

compression = "d", quote = "Close")

rem<- get.hist.quote(instrument = "rem", start = date ,

compression = "d", quote = "Close")

fty<- get.hist.quote(instrument = "fty", start = date ,

compression = "d", quote = "Close")

gnr<- get.hist.quote(instrument = "gnr", start = date ,

compression = "d", quote = "Close")

#### calculate correlations and plot network##################

x <- merge(acwi,vwo,iwv,mcro,ief,igov,pcy,hdg,rem,fty,gnr)

x<- na.omit(x)
colnames(x)<- c("acwi","vwo", "ivw", "mcro", "ief", "igov", "pcy", "hdg", "rem", "fty", "gnr" )

returns =diff(x)/x[-length(x)] # Calculating rates of return from price levels

returns = na.omit(returns)

cor=cor(returns)

qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout= "spring",vsize=10)

Figure 6 3
As above but replace last line with:
qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout= "spring",vsize=10,minimum=0.5)

Figure 6 5
As above but replace last line with:
qgraph(cor,shape="circle",posCol="darkgreen",negCol="darkred",layout= "circular",vsize=10)
Further resources
Further reading on the packages mentioned in this chapter see:
Adrian Trapletti and Kurt Hornik (2013). tseries: Time Series Analysis and
Computational Finance. R package version 0.10-32.

Diethelm Wuertz and many others (2013). fImport: Rmetrics - Economic and
Financial Data Import. R package version 3000.82. http://CRAN.R-
project.org/package=fImport

Peter Carl and Brian G. Peterson (2013). PerformanceAnalytics: Econometric

tools for performance and risk analysis.. R package version 1.1.0.
http://CRAN.R-project.org/package=PerformanceAnalytics

Sacha Epskamp, Angelique O. J. Cramer, Lourens J. Waldorp, Verena D.

Schmittmann, Denny Borsboom (2012). qgraph: Network Visualizations of
Relationships in Psychometric Data. Journal of
Statistical Software, 48(4), 1-18. URL http://www.jstatsoft.org/v48/i04/.

For further information on the datasets mentioned in this chapter see:

Gao X, An H, Zhong W (2013) Features of the Correlation Structure of Price
Indices. PLoS ONE 8(4): e61091. doi:10.1371/journal.pone.0061091
Grossman, Adam D., et al. "Altering physiological networks using drugs: steps
towards personalized physiology." BMC Medical Genomics 6.2 (2013): 1-11.
Taylor, Véronique A., et al. "Impact of meditation training on the default mode
network during a restful state." Social cognitive and affective neuroscience 8.1
(2013): 4-14.
Wtite you notes below…
 7. How to use a Rose for directional impact

Figure 7 1 Rose diagram of beaver foraging

Directional datasets were first analyzed in the Earth Sciences, Meteorology and
Medicine. With the growth in social computing and growing volumes of data
from online sources, it will become more frequently observed in nontraditional
areas also. The analysis of this type of data poses a number of challenges for the
data analyst. For example, the usual statistics such as the arithmetic mean or
standard deviation are not rotational invariant. For example, the average
direction between 10 degrees and 350 degrees is not given by the arithmetic
mean of 180 degrees. However, Rose diagrams offer a powerful solution to the
issue of presenting circular or directional data. Figure 7 1 shows the foraging
direction of three beavers given in Lewis (2013). Each beaver is represented by
a different color (orange, red, blue). The foraging directional preferences of
each beaver are clearly discernible from the diagram.
Figure 7 2 Pollution rose of Carbon monoxide concentration at Marylebone

Figure 7 2 illustrates a slight variation of the basic rose diagram know as a

pollution rose. It plots the frequency of counts by wind direction for carbon
monoxide concentration (parts per million) measured at Marylebone in central
London. The petals are similar to the bars in a histogram and represent the
frequency of observations. In this image the petals are color coded to capture
the extent of concentration. Circular bands are drawn at 5%,10%,15% and
20%. A petal that touches the 5% band represents 5% of the overall number of
observations.
Figure 7 3 Australian wind rose

Figure 7 3 illustrates another variation of the rose diagram - the Australian

wind rose. The chart shows a traditional Australian wind rose using data
measured at Marylebone, London between 1st January 1998 to 23rd June 2005.
Wind directions are divided into eight compass directions. The circles around
the image represent the various percentages of occurrence of the winds. For
example, the branch to the north west just reaches the 10% ring implying a
frequency of 10% blowing from that direction. Calm has no direction.
Innovative use of rose diagrams
Florence Nightingale: A British Social Reformer, Statistician, and founder of
modern nursing. She made use of the rose diagram during the late 1850’s and
early 1860’s to convince politicians that deaths from epidemic disease were
avoidable. Her evidence presented in the g-graphic below, was
compelling[viii].

Fault patterns: Leonard, and Szynkaruk (2013) investigate the fault pattern in
areas of the Polish Outer Carpathians. Fault positions were established by
photointerpretation where zones of breccia and cataclasites extended on valley
slopes as narrow and often dry gullies. The researchers used rose diagrams to
visualize the distribution of fault pattern.
Storm surges: The South China Sea (SCS) lies under the southwest-northeast
pathway of the seasonal monsoons which dominate the large scale sea level
dynamics of region. Tkalich, Pavel, et al (2013) investigate wind-driven sea
level anomalies using tide gauge data and satellite altimetry. Climatological
wind rose diagrams using data over the period 1984 – 2007 were used as a
powerful visualization technique.
Creating rose diagrams in R
Figure 7 1 was created using the circular package. It was built using the following
three steps:
Step 1: Download and install the required packages.
install.packages("circular")
require(circular)

Step 2: Clean and prepare the data

set.seed(1234)
beaver.1 <- rvonmises(n=50, mu=circular(0), kappa=3)
beaver.2 <- rvonmises(n=50, mu=circular(pi), kappa=3)
beaver.3 <- rvonmises(n=50, mu=circular(pi/2), kappa=3)

Step 3: Plot the chart

par(bg="lightgray")
rose.diag(beaver.1, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkred",tol=0,prop=1.5)
par(new=T)
rose.diag(beaver.2, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkblue",tol=0,prop=1.5)
par(new=T)
rose.diag(beaver.3, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkorange",tol=0,prop=1.5)
points(beaver.1, col="darkred", stack=TRUE)
points(beaver.2, col="darkblue", stack=TRUE)
points(beaver.3, col="darkorange", stack=TRUE)

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Creates the required data. In this case the beaver foraging directions are
actually simulated from a Von mises distribution. The argument set.seed(1234)
ensures you can replicate exactly the chart as it is presented here. Note the
foraging directions are stored in beaver.1 , beaver.2 and beaver.3 respectively
Step 3: The rose.diag function draws the image. The trick is to overlay three rose
diagrams. This is achieved using the function par(new=T) before the second and
third image are drawn. The actual observations for each beaver are plotted in
their respective colors on the diagram using the function points .
Tip: The background color of the image is controlled by par(bg="lightgray"). You can
set to an alternative color by specifying your choice. A popular option
is par(bg="cornsilk") .

The pollution rose of Figure 7 2 was created using the openair package. It was
constructed using the following three steps:
Step 1: Download and install the required packages.
install.packages("openair")
require(openair)

Step 2: Clean and prepare the data

data(mydata)
pollutant <- c("co")# Carbon monoxide data

Step 3: Plot the chart

title=c("Carbon monoxide concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")
pollutionRose(mydata,paddle=FALSE,pollutant= pollutant,annotate=F, key.footer = pollutant, key.position
= "right", key = TRUE, breaks = 8,main=title, sub=sub,auto.text=FALSE)

Step 1: Installs the required packaged and makes it available to R.

Step 2: Loads the dataframe which amongst other items, contains data on
carbon monoxide concentration, in ppm, as a numeric vector. The carbon
monoxide data is assigned to the variable pollutant.
Step 3: The pollutionRose function draws the image.
Tip 1: The location of the key is determined by the argument key.posiition ; It
can take values “left”,”right”,”top” and “bottom”. The number of colors (or
breaks) represented in the key is determined by the argument breaks; for many
datasets values between 6-8 work well.

Tip 2: You can change the plot to a modern wind rose by setting the
argument paddle=TRUE .

Tip 3: When substituting your own data note the pollutionRose function
requires data frames with a field "date" that can be in either POSIXct or Date
format but should be set to the Greenwich Mean Time time-zone. The data
should be at least hourly or higher resolution.
Tip 4: The chart used the argument pollutant <- c("co") to access the carbon
monoxide observations in the mydata data frame. This data frame contains a
number of other air quality observations taken in central London:

1. "ws" - Wind speed.

2. "wd" - Wind direction, in degrees from North.
3. "nox"- Oxides of nitrogen concentration.
4. "no2" - Nitrogen dioxide concentration.
5. "o3"- Ozone concentration.
6. "pm10" - Particulate PM10 fraction measurement.
7. "so2" - Sulfur dioxide concentration.
8. "pm25" - Particulate PM2.5 fraction measurement.
The Australian wind rose, Figure 7 3 , was created using the plotrix package. It
was constructed using the following three steps:
Step 1: Download and install the required packages.
install.packages("openair")# contains the data
install.packages("plotrix")# used to draw the chart
require(openair)
require(plotrix)

Step 2: Clean and prepare the data

raw.data<-cbind(mydata$wd,mydata$ws)
data<-na.omit(raw.data)
table<-bin.wind.records(data[,1],data[,2]*3.6,ndir=8)#

Step 3: Plot the chart

par(bg="cornsilk")
oz.windrose(table)

Step 1: Installs the required packages and makes them available to R.

Step 2: Creates a variable called raw.data which contains data on wind speed
and wind direction. Missing observations are removed using na.omit(raw.data). The
function bin.wind.records creates a table and covert wind speed to km per hour.
Step 3: The oz.windrose function draws the image.
Tip 1:Data for the oz.windrose function needs to be in a format where the rows
represent speed ranges and the columns indicating wind directions.
Furthermore, the data should be in percentages such that the sum of all data in
your table is equal to 100. This is best achieved using the bin.wind.records
function. This function classifies wind direction and speed records into a matrix
of percentages of observations in speed and direction bins.Use the
argument ndir to change the number of direction bins to be used in the wind
rose. The traditional number of bins is ndir=8 .
Tip 2:Use the argument speed.col in the oz.windrose function to change the
color of the branches of the wind rose. The default colors, shown in the above
chart are obtained by speed.col
=c("#dab286","#fe9a66","#ce6733","#986434") . You could also set
alternative colors such as speed.col =c("red","green","purple","yellow") .
Tip 3:To remove the legend add the argument show.legend=FALSE to
the oz.windrose function.
Summary
Directional datasets were first analyzed in the Earth Sciences, Meteorology and
Medicine. Rose diagrams proved an excellent tool for visualizing directional
data. Similar to a histogram they provide a quick and simply technique for
investigating the distribution of circular datasets. With the growth in social
computing and growing volumes of data from online sources, Rose diagrams are
becoming more frequently used in a wide range of ‘non-traditional’ areas also.
Quick reference R code
Figure 7 1 Rose diagram of beaver foraging
install.packages("circular")
require(circular)
set.seed(1234)
beaver.1 <- rvonmises(n=50, mu=circular(0), kappa=3)
beaver.2 <- rvonmises(n=50, mu=circular(pi), kappa=3)
beaver.3 <- rvonmises(n=50, mu=circular(pi/2), kappa=3)
par(bg="lightgray")
rose.diag(beaver.1, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkred",tol=0,prop=1.5)
par(new=T)
rose.diag(beaver.2, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkblue",tol=0,prop=1.5)
par(new=T)
rose.diag(beaver.3, bins=18,axes=TRUE,ticks=TRUE,border="black",col="darkorange",tol=0,prop=1.5)
points(beaver.1, col="darkred", stack=TRUE)
points(beaver.2, col="darkblue", stack=TRUE)
points(beaver.3, col="darkorange", stack=TRUE)

Figure 7 2 Pollution rose of Carbon monoxide concentration at Marylebone

install.packages("openair")
require(openair)
data(mydata) pollutant <- c("co")title=c("Carbon monoxide concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")
pollutionRose(mydata,paddle=FALSE,pollutant= pollutant,annotate=F, key.footer = pollutant, key.position
= "right", key = TRUE, breaks = 8,main=title, sub=sub,auto.text=FALSE)

Figure 7 3 Australian wind rose

install.packages("openair")
install.packages("plotrix")
require(openair)
require(plotrix)
raw.data<-cbind(mydata$wd,mydata$ws)
data<-na.omit(raw.data)# remove missing observations
table<-bin.wind.records(data[,1],data[,2]*3.6,ndir=8)# create table and covert wind speed to km per
hour
par(bg="cornsilk")
oz.windrose(table)
Further resources
Further reading on the packages mentioned in this chapter see:
C. Agostinelli and U. Lund (2011). R package 'circular': Circular
Statistics (version 0.4-3). URL https://r-forge.r-project.org/projects/circular/
Carslaw, D.C. and K. Ropkins, (2012) openair --- an R package for air
quality data analysis. Environmental Modelling & Software. Volume
27-28, 52-61.
Lemon, J. (2006) Plotrix: a package in the red light district of R.
R-News, 6(4): 8-12.
For further information on the datasets mentioned in this chapter see:
Lewis, N.D. (2013). The Visual Display of Quantitative Data. Heather Hills
Press.
Mastella, Leonard, and Ewa Szynkaruk. "Analysis of the fault pattern in selected
areas of the Polish Outer Carpathians." (2013): 263-276.
Tkalich, Pavel, et al. "Storm surges in the Singapore Strait due to winds in
the South China Sea." Natural Hazards (2013): 1-18.
Wtite you notes below…
 8. Maximize trends in multivariate timeseries
data

Figure 8 1 Multivariate shaded timeseries plot of European stock markets

Multivariate shaded timeseries plots allow easy identification of relationships

between numerous variables and trends in those variables individually and
simultaneously across time. Figure 8 1 illustrates such a plot for the 90 day
change in stock prices for four European stock market indices (FTSE,CAC, SMI
and DAX) using data collected over the period 1991-1998. Notice how the
shading of colors ranges from dark purple for negative 90 changes in price, to
deep green for very positive changes in price. Observe also how the image
allows you to identify the trend in price changes across time and also between
stock markets. In this way the multivariate shaded timeseries plot maximizes the
visibility of any trends within the sample data.
Innovative use of the multivariate shaded timeseries plot
One of the major uses of a multivariate shaded timeseries plot is to visualize
trends between and within variables across time. This is an extremely useful
property for exploratory data analysis of datasets which contain large numbers
of timeseries data. This chart should be a key technique in your data
visualization toolbox.
Sunspots analysis: Sunspots are cool planet-sized areas where intense
magnetic loops poke through the Sun’s visible surface. They are measured by
scientists and used to track solar cycles. Astronomers have been monitoring and
collecting data on them since the 1740’s. An interesting question is the degree to
which sunspot activity changes by month of the year and also across time. This
ides is explored in Figure 8 2 , which shows sunspot activity from the mid-18th
century to the end of the end of the 20th century. The trend in sunspot activity
across time and month is clearly evident, even with this relatively large dataset.
The deeper the green the higher is sunspot activity. Low sunspot activity is
reflected in deep purple. The relationship between sunspot activity across
months and across time is quite striking. The line chart at the bottom of the
diagram reports median activity across all months.

Figure 8 2 Sunspot activity by month of year

Creating multivariate shaded timeseries plots in R
Figure 8 1 was created using the mvtsplot package. It was built using the following
three steps:
Step 1: Download and install the required packages.
install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)

Step 2: Clean and prepare the data

data <- diff(EuStockMarkets ,90)

Step 3: Plot the chart

mvtsplot(data,norm ="internal", levels = 5,margin=FALSE)

Step 1: Installs the required packaged and makes it available to R. If the

package is already installed on your machine you do not need to
use install.packages . However, the require command is needed at the start of a new
R session.
Step 2: Loads the required data into the R environment. The EuStockMarkets
dataframe is contained in the datasets package/ It is pre-installed in the basic R
package. The 90 day change in price level is calculated and stored in the
variable called data.
Step 3: The mvtsplot function draws the image.
Tip 1: The parameter levels = 5 patriciansthe data into five levels. Each level is
assigned a particular shade of purple or green. Increase the value assigned
to levels to increase the range of shading. Figure 8 3 illustrates this with levels =
2 for the Europeanstock market data of Figure 8 1 .

Tip 2: The parameter norm ="internal" ensures each individual time series is
categorized into colors based on the its range of values. This implies the same
color in two different time series may have a different interpretations. If "global
“is specified, then each time series will be categorized based on the range of
values for the entire collection of time series.
Figure 8 3 Shaded timeseries plot of European stock markets with levels =2
Summary
Multivariate shaded timeseries plots allow easy identification of relationships
between numerous variables and trends in those variables individually and
simultaneously across time. This is an extremely useful property for exploratory
data analysis of datasets which contain large numbers of timeseries data. This
chart should be frequently deployed for exploratory analysis and is an important
technique in your data visualization and presentation toolkit.
Quick reference R code
Figure 8 1 Multivariate shaded timeseries plot of European stock markets
install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data <- diff(EuStockMarkets ,90)
mvtsplot(data,norm ="internal", levels = 5,margin=FALSE)

Figure 8 2 Sunspot activity by month of year

install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data<-matrix(data = sunspot.month, ncol = 12, byrow = TRUE)
colnames(data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul", "Aug","Sep","Oct","Nov","Dec")
rownames(data)<-(1749:1997)
time<-(1749:1997)
mvtsplot(data,xtime = time,norm ="global", levels = 5)

Figure 8 3 Shaded timeseries plot of European stock markets with levels =2

install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data <- diff(EuStockMarkets ,90)
mvtsplot(data,norm ="internal", levels = 2,margin=FALSE)
Further resources
Further reading on the mvtsplot package can be found in:
Roger D. Peng (2012). mvtsplot: Multivariate Time Series Plot. R
package version 1.0-1. http://CRAN.R-project.org/package=mvtsplot.
For further details on the datasets mentioned in this chapter see:
R Core Team (2013). R: A language and environment for statistical
computing. R Foundation for Statistical Computing, Vienna, Austria.
URL http://www.R-project.org/.
Wtite you notes below…
 PART 2: FIFTY VISULIZATIONS
TO UNLEASH YOUR
CREATIVITY
This section of the book shows you how to create fifty scientific data
visualizations easily and quickly, using the free and widely available R
computing package. It contains no theory, no complex equations, you don’t even
need to have prior experience. Select the chart of interest, type in the example
and you are done. All charts in this section can be reproduced by you in ten
minutes or less. Once you are comfortable with a particular data visualization,
feel free to adapt it for your own purposes. Tips on how to do this are discussed
throughout the text. Each chart is presented alongside step by step instructions
on how to quickly construct it. This is followed by practical tips to enhance the
visual experience. Finally, at the end of each chart the entire code used to create
the chart is given for your reference. If you are in a hurry just copy out this
section into R, make changes for your data and you are done!
 9. Tips on Color and Shapes in R
Let’s work through an example. If this were a computer programming book, we
would begin with a program to produce the statement “Hello World!”; since this
is text about data visualization, let us begin with a random circle plot. Here is
the code, type it directly into R:.
#random circles plot
set.seed(123) # set random number to known value
n = 5# number of circles to draw
x = runif(n) #generate n random numbers
size=abs(x)/sqrt(pi) # set the size of the circles
symbols(x,circles=size) # plot the circles

The first line allows you to replicate exactly the image below. It basically starts
the random number generator at the same value every time. The second line
indicates five circles will be drawn (to draw more set n to a larger number).
The third line generates the random numbers. The fifth line sets the size of the
circles as a function of the random number value. The final line draws the
random circles image onto the R graphics device. You should now see on your
R graphics device an image that looks like the one below:
Not that impressive but at least you have drawn something! Let’s play a little
with it. First let’s make the circumference of the circles blue; to do this type:
symbols(x,circles=size,fg="blue")# turn circumference blue
You should now see an image that looks like this:
-
The argument fg changes the color of the circumference of the circle. R has a
rich array of color options. To get a list of the basic color type colors() into the
R consul and press return.
Back to our example. We would like to fill the circles with the color green. This
can be accomplished in one line
symbols(x,circles=size,fg="blue",bg="green")#green circles
The resultant image is shown below:

Since the circles are rather large, it might be a good idea to limit their size. This
can be done by adding the inches argument to the symbols function. We will set
the maximum size to 0.5 of an inch and set the background color of the image to
cornsilk. This is achieve using the par() function.
par(bg="cornsilk")# set background color
symbols(x,circles=size,fg="blue",bg="green", inches=0.5)
The image below is the result:
Of course, we will need to label the x and y axis and give the image a title. This
can achieved by typing one line into R:
symbols(x,circles=size,fg="blue",bg="green", inches=0.2, main="Random
Circle Chart", xlab="Five random bubbles",ylab="Bubble size")# random
circles chart finished!
The diagram below is the final product:
In reality it only took two lines of code to produce this image:
par(bg="cornsilk")# set background color
symbols(x,circles=size,fg="blue",bg="green", inches=0.2, main="Random
Circle Chart", xlab="Five random bubbles",ylab="Bubble size")# draw chart
Not bad for a couple of lines! But we can do even better when we visualize real
data in subsequent sections.
Tip 1: To get help on a function, type help(function name) or ? function name.
For example to get help on the par function you can type:
help(par)
or
?par
A note on color in R
Traditionally scientific charts have used color sparingly. This has begun to
change as more and more researchers become aware of both the power of color
and the ease by which it can be added to an image. R is particularly rich in
color. Let’s try some of the built in palettes right now! We begin with the
rainbow palette:

The heat.colors palette is great for heat maps:

As is the terrain palette for geographical maps and topography:

The topo.colors palette are a little brighter:

You can even cool things down with cm.colors:

We have just scratched the surface. Creative ways to use these and other colors
will be explored in the next sections of this text.
Now let’s get started!
 10. Bubble plot

Figure 10 1 Lifecycle savings for a range of countries

How to read the chart

The chart displays the relationship between personal saving (x-axis), growth
rate in disposable income (y-axis) and proportion of population less than 15
years old (size of bubble).

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

require(datasets)# contains the data
require (graphics)# used to draw the plot
attach(LifeCycleSavings)

Step 2: Clean and prepare the data

# identify the nations to use in the chart
nations<-c("Chile","Denmark","Greece", "Iceland",
"Japan","Korea","Norway","Netherlands","United Kingdom", "Jamaica",
"Libya" )
# store nations data in a data frame called “data”
data<- LifeCycleSavings [match(nations, row.names(LifeCycleSavings)), ]
Step 3: Plot the chart
par(bg="white")# set the background color
# plot the axes and axes labels
plot(0:23, 0:23, type="n", xlab="Personal savings rate",ylab="Growth rate of
real per-capita disposable income",frame.plot=F)
# plot the orange bubbles
symbols(data$sr, data$ddpi, circles = sqrt(data$pop15/pi),
fg = "white", inches=0.25,bg="orange", add=T)
#add country names
text(data$sr, data$ddpi, rownames(data), cex=0.7,adj=c(-0.5,0.6),col="black")
#add title
title("Intercountry Life-Cycle Savings",col.main="black",line=2, cex.main =2)
title("1960–1970",col.main="black",line=1, cex.main =1)

Visualization quick tips

Tip 1: Change the background color of the plot by using the argument
par(bg="my color"). The color of the bubbles is controlled by the argument bg
in the symbols functions; you can also change the border around the bubbles by
setting argument fg="my color".

Tip 2: The radius of the bubbles is controlled by the circles argument in the
symbols function. The chart used circles = sqrt(data$pop15/pi). To ensure the
bubbles are not to large the maximum size of each bubble was limited by setting
inches=0.25. For larger bubbles increase this number, for smaller bubbles
decrease the number, the default is 1 inch.

Tip 3: The title function offers considerable flexible in adding text to a chart.
The color of the text can be changed by setting col.main ="my color". The
argument cex.main controls the size of the text. The bubble chart used a title,
with cex.main=2, and a subtitle with a smaller font size set by cex.main =1.
Quick reference R code

require(datasets)# contains the data

require (graphics)# used to draw the plot
attach(LifeCycleSavings)
nations<-c("Chile","Denmark","Greece", "Iceland",
"Japan","Korea","Norway","Netherlands","United Kingdom", "Jamaica",
"Libya" )

data<- LifeCycleSavings [match(nations, row.names(LifeCycleSavings)), ]

par(bg="white")
plot(0:23, 0:23, type="n", xlab="Personal savings rate",ylab="Growth rate of
real per-capita disposable income",frame.plot=F)
symbols(data$sr, data$ddpi, circles = sqrt(data$pop15/pi),
fg = "white", inches=0.25,bg="orange", add=T)
text(data$sr, data$ddpi, rownames(data), cex=0.7,adj=c(-0.5,0.6),col="black")
title("Intercountry Life-Cycle Savings",col.main="black",line=2, cex.main =2)
title("1960–1970",col.main="black",line=1, cex.main =1)
 11. Multivariate shaded time series plot

Figure 11 1 The 90 day price change of four European stock market indices

How to read the chart

The chart is constructed from daily data on the four major European stock
market indices (FTSE, CAC, SMI and DAX) recorded over the period 1991 to
1998. The panel shows the time series for the 90 day price changes for each
market index. The values have been discretized and assigned to five distinct
categories. The shading at the extremes represents whether a high 90 day change
(green) or low 90 day change (purple) in prices were observed. The deeper the
purple, the lower the 90 day price change. The deeper the green, the higher the
90 day price change.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("mvtsplot")# you only need to do this once
install.packages("datasets") # you only need to do this once
require(datasets) # package which contains the data
require(mvtsplot) # loads chart drawing package into R
Step 2: Clean and prepare the data
data <- diff(EuStockMarkets ,90)
# The data are contained in the EuStockMarkets data frame. The function diff()
calculates the 90 day difference. The cleaned observations are stored in the
variable called data.

Step 3: Plot the chart

mvtsplot(data,norm ="internal", levels = 5,margin=FALSE)

Visualization quick tips

Tip 1: The mvtsplot function requires data in a matrix format. When using your
own data remember to set the argument ncol equal to the number of columns in
your data set.

Tip 2: You can supply your own column heading by setting colnames<-(data
c("name.1","name.2","name.3",…). Again, make sure you have sufficient
column names for the number of columns in your data.

Tip 3: The argument norm in the mvtsplot function addresses whether the
categorization of the time series is based on within-series categories (norm =
"internal") or using all the data in (norm = "global"). In the chart created in this
chapter the categories were the four stock market indices (FTSE, CAC, SMI and
DAX) with each categorization created using within-series data. Experiment
with both norm = "internal" and norm = "global" to see if trends in the overall
dataset appear more clearly.

Tip 4: The argument levels in the mvtsplot function controls how many colors
are used in the plot. The default is levels =3. In the chart created in this chapter
the values of each time series are divided into five categories ("very low",
"low", "medium", "high" and "very high") and plotted using deep purple for very
low values and deep green to represent high values. As a rule of thumb smooth
data can usually be plotted with a larger number of levels; very noisy or spikey
data typically needs to be plotted with fewer levels. Try experimenting with
different values to see if trends in the overall dataset appear more clearly.

Quick reference R code

install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data <- diff(EuStockMarkets ,90
mvtsplot(data,norm ="internal", levels = 5,margin=FALSE)
NOTES

Write your notes below…

 12. Multivariate marginal shaded time series
plot

Figure 12 1 Monthly sunspots recorded over the period 1749 to 1997

How to read the chart

The chart is constructed from data on the monthly number of sunspots recorded
over the period 1749 to 1997. The main panel shows the time series of sunspots
for each month of the year. The values of each time series are discretized and
assigned to distinct categories. The shading represents whether a high number
(green) or low number (purple) of sunspots were recorded. The right panel
contains box plots for each of the twelve months. The median value is
represented by a large dot. The lower panel presents a time series plot of the
median value of observations for each year.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("mvtsplot")# you only need to do this once
install.packages("datasets") # you only need to do this once
require(datasets) # package which contains the data
require(mvtsplot) # loads chart drawing package into R
Step 2: Clean and prepare the data
data<-matrix(data = sunspot.month, ncol = 12, byrow = TRUE)
# The sunspots data is contained in the data-frame sunspot.month. It contains
monthly observations from 1749 to 1997. Since there are 12 months in the year
the number of columns is set equal to 12 (ncol=12). The parameter byrow is set
to TRUE because the data in sunspot.month is stored in rows. The cleaned
observations are stored in the variable called data.

colnames(data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul",

"Aug","Sep","Oct","Nov","Dec")# Add column names for each month of the
year.

rownames(data)<-(1749:1997) # Data cover the years 1749 to 1997

time<-(1749:1997) # We use this variable to plot years 1749 to 1997 in the
lower panel.

Step 3: Plot the chart

mvtsplot(data,xtime = time,norm ="global", levels = 2)

Visualization quick tips

Tip 1: The mvtsplot function requires data in a matrix format. When using your
own data remember to set the argument ncol equal to the number of columns in
your data set.

Tip 2: You can supply your own column heading by setting colnames<-(data
c("name.1","name.2","name.3",…). Again, make sure you have sufficient
column names for the number of columns in your data.

Tip 3: The argument norm in the mvtsplot function addresses whether the
categorization of the time series is based on within-series categories (norm =
"internal") or using all the data in (norm = "global"). In the chart created in this
chapter the categories were months of the year using all the data. Experiment
with both norm = "internal" and norm = "global" to see if trends in the overall
dataset appear more clearly.

Tip 4: The argument levels in the mvtsplot function controls how many colors
are used in the plot. The default is levels =3. In this case the values of each time
series are divided into three categories ("low", "medium", and "high") and
plotted using purple for low values, grey for medium values, and green to
represent high values. As a rule of thumb smooth data can usually be plotted
with a larger number of levels; very noisy or spikey data typically needs to be
plotted with fewer levels. Try experimenting with different values to see if
trends in the overall dataset appear more clearly.

Quick reference R code

install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data<-matrix(data = sunspot.month, ncol = 12, byrow = TRUE)
colnames(data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul",
"Aug","Sep","Oct","Nov","Dec")
rownames(data)<-(1749:1997)
time<-(1749:1997)
mvtsplot(data,xtime = time,norm ="global", levels = 2)
NOTES

Write your notes below…

 13. Multivariate split marginal shaded time
series plot

Figure 13 1 Monthly sunspot data versus random log-normal data

How to read the chart

The chart is constructed from data on the monthly number of sunspots recorded
over the period 1749 to 1997. The main panel is divided into two sections by a
red line. The section above the red line shows the time series of sunspots for
each month of the year. The area below the red line presents simulated data of
sunspots from a lognormal distribution. The values of both the actual and
simulated time series are discretized and assigned to distinct categories. The
shading represents whether a high number (green) or low number (purple) of
sunspots were recorded. The right panel contains box plots for each series for
each month of the year. The median value is represented by a large dot. The
lower panel presents a time series plot of the median value of both sets of
observations by year.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("mvtsplot")# you only need to do this once
install.packages("datasets") # you only need to do this once
require(datasets) # package which contains the data
require(mvtsplot) # loads chart drawing package into R

Step 2: Clean and prepare the data

data<-matrix(data = sunspot.month, ncol = 12, byrow = TRUE)
# transform raw data in to matrix format
colnames(data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul",
"Aug","Sep","Oct","Nov","Dec")# Add labels for the months of the year.
rownames(data)<-(1749:1997)# Data cover the years 1749 to 1997.
time<-(1749:1997)# Data cover the years 1749 to 1997.
set.seed(1234)# this allows you to replicate the chart.
random.data<- matrix(rlnorm (12*nrow(data),mean=3.9,sd=1.2),
nrow(data),12) # Generate log-normal random variables.
colnames(random.data) <- c("Jan","Feb","Mar","Apr","May" ,"Jun", "Jul",
"Aug","Sep","Oct","Nov","Dec")
# Give the log-normal data columns a heading representing months of the year.
group<-c(0,0,0,0,0,0, 0,0,0,0,0,0,1,1,1,1,1,1, 1,1,1,1,1,1)
# group is used to separate the sunspot data from the random data.
data<-cbind(random.data,data)
# cbind is a function which combines the actual sunspot data with the simulated
data.

Step 3: Plot the chart

mvtsplot(data,xtime = time, norm ="internal", levels = 3,

group=group,gcol="red")

Visualization quick tips

Tip 1: The mvtsplot function requires data in a matrix format. When using your
own data remember to set the argument ncol equal to the number of columns in
your data set.

Tip 2: You can supply your own column heading by setting colnames<-(data
c("name.1","name.2","name.3",…). Again, make sure you have sufficient
column names for the number of columns in your data.

Tip 3: The argument norm in the mvtsplot function addresses whether the
categorization of the time series is based on within-series categories (norm =
"internal") or using all the data in (norm = "global"). Experiment with both norm
= "internal" and norm = "global" to see if trends in the overall dataset appear
more clearly.

Tip 4: The argument levels in the mvtsplot function controls how many colors
are used in the plot. The default is levels =3 which was used in the chart in this
chapter. For the default value purple is assigned to low values, grey to medium
values, and green to high values. As a rule of thumb smooth data can usually be
plotted with a larger number of levels; very noisy or spikey data typically needs
to be plotted with fewer levels. Try experimenting with different values to see if
trends in the overall dataset appear more clearly.

Quick reference R code

install.packages("mvtsplot")
install.packages("datasets")
require(datasets)
require(mvtsplot)
data<-matrix(data = sunspot.month, ncol = 12, byrow = TRUE)
colnames(data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul",
"Aug","Sep","Oct","Nov","Dec")
rownames(data)<-(1749:1997)
time<-(1749:1997)
set.seed(1234)
random.data<- matrix(rlnorm (12*nrow(data),mean=3.9,sd=1.2),
nrow(data),12)
colnames(random.data) <- c("Jan","Feb","Mar","Apr","May","Jun","Jul",
"Aug","Sep","Oct","Nov","Dec")#
group<-c(0,0,0,0,0,0, 0,0,0,0,0,0,1,1,1,1,1,1, 1,1,1,1,1,1)
data<-cbind(random.data,data)
mvtsplot(data, xtime = time, norm ="internal", levels = 3, group=group,
gcol="red")
NOTES

Write your notes below…

 14. Word cloud

Figure 14 1 United States State of the Union Addresses (2010 and 2011)

How to read the image

The image displays the most frequently used words by President Obama in his
State of the Union addresses during 2010 and 2011. Word frequency is
visualized by the size of the text; the more frequent the word the-e larger it
appears in the image.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)

Step 2: Clean and prepare the data

data(SOTU)# contains the text of the Presidential addresses.
# we only want the words so we remove punctuation
text <- tm_map(SOTU, removePunctuation)
# remove standard stop words. These are standard words which are removed
prior to processing.
text <- tm_map(text, function(x)removeWords (x,stopwords()))
# put cleaned data in appropriate format
tdm <- TermDocumentMatrix(text)
m <- as.matrix(tdm)
v <- sort(rowSums(m),decreasing=TRUE)
d <- data.frame(word = names(v),freq=v)
Step 3: Plot the chart
par(bg="black")# set background color
wordcloud(d$word,d$freq, random.order=FALSE,min.freq=5 ,
color="gold")points(fly.2, plot.info=res, col="darkblue", stack=TRUE)

Visualization quick tips

Tip 1: To create a word cloud your data needs to be in a corpus format. One
way to do this is to transform your text into a data frame and then to a corpus (a
collection of text documents). Something along the lines of
library(tm)
textframe <- do.call("rbind", lapply(mytext, as.data.frame))
myCorpus <- Corpus(VectorSource(textframe$text))
Note VectorSource requires character vectors.

Tip 2: The argument min.freq controls the minimum word count displayed in the
image. Words which occur less than min.freq are excluded from the word cloud.

Tip 3: To limit the maximum number of words in the image add the max.words
argument to the wordcloud function i.e. wordcloud(max.word=100,…).

Tip 4: To change the color of the printed text use the argument colors =
“my_favorite_color”.

Quick reference R code

install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)
data(SOTU)
text <- tm_map(SOTU, removePunctuation)
text <- tm_map(text, function(x)removeWords(x,stopwords()))
tdm <- TermDocumentMatrix(text)
m <- as.matrix(tdm)
v <- sort(rowSums(m),decreasing=TRUE)
d <- data.frame(word = names(v),freq=v)
par(bg="black")
wordcloud(d$word,d$freq, random.order=FALSE,min.freq=5 , color="gold")
NOTES

Write your notes below…

*
 15. Commonality word cloud

Figure 15 1 United States State of the Union Addresses (2010 and 2011)

How to read the image

The image displays high frequently commonality words used by President

Obama in his State of the Union addresses during 2010 and 2011. Word
commonality is visualized by the size of the text; the greater the commonality of
a word the larger it appears in the image.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)

Step 2: Clean and prepare the data

data(SOTU)# contains the text of the Presidential addresses.
text <- SOTU
#below the data is cleaned
text <- tm_map(text, removePunctuation)
text <- tm_map(text, tolower)
text <- tm_map(text, removeNumbers)
# remove standard stop words. These are standard words which are removed
prior to processing.
text <- tm_map(text, function(x)removeWords(x,stopwords()))
# put cleaned data in appropriate format
term.matrix <- TermDocumentMatrix(text)
term.matrix <- as.matrix(term.matrix)

Step 3: Plot the chart

par(bg="yellow")# set chart background color to yellow.
set.seed(123)# use to allow the replication of the image
commonality.cloud(term.matrix,random.order=FALSE)

Visualization quick tips

Tip 1: To create a commonality word cloud your data needs to be in a corpus

format. One way to do this is to transform your text into a data frame and then to
a corpus (a collection of text documents). Your R code with look similar to the
text below
library(tm)
textframe <- do.call("rbind", lapply(mytext, as.data.frame))
myCorpus <- Corpus(VectorSource(textframe$text))
Note VectorSource requires the source be character vectors.

Tip 2: To limit the maximum number of words in the image add the max.words
argument to the commonality.cloud function i.e. commonality.cloud
(max.word=100,…).

Tip 3: To see the list of stop words used type the following into the R console
window: stopwords("en")
Quick reference R code
install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)
data(SOTU)
text <- SOTU
text <- tm_map(text, removePunctuation)
text <- tm_map(text, tolower)
text <- tm_map(text, removeNumbers)
text <- tm_map(text, function(x)removeWords(x,stopwords()))
term.matrix <- TermDocumentMatrix(text)
term.matrix <- as.matrix(term.matrix)

par(bg="yellow")
set.seed(123)
commonality.cloud(term.matrix,random.order=FALSE)
NOTES

Write your notes below…

 16. Comparison word cloud

Figure 16 1 United States State of the Union Addresses (2010 and 2011)

How to read the image

The image displays high frequently words used by President Obama in his State
of the Union addresses during 2010 and 2011. The top half of the diagram are
frequently used word from the 2010 address (green), the bottom half are high
frequency words from the 2011 address (burnt orange).

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)

Step 2: Clean and prepare the data

data(SOTU)# contains the text of the Presidential addresses.
text <- SOTU
#below the data is cleaned
text <- tm_map(text, removePunctuation)
text <- tm_map(text, tolower)
text <- tm_map(text, removeNumbers)
# remove standard stop words. These are standard words which are removed
prior to processing.
text <- tm_map(text, function(x)removeWords(x,stopwords()))

# put cleaned data in appropriate format

term.matrix <- TermDocumentMatrix(text)
term.matrix <- as.matrix(term.matrix)
colnames(term.matrix) <- c("2010 Presidential Address","2011 Presidential
Address")

Step 3: Plot the chart

set.seed(1234)
par(bg="gray90")# set background color to gray.
comparison.cloud(term.matrix,max.words=300,random.order=FALSE)

Visualization quick tips

Tip 1: To create a comparison word cloud your data needs to be in a corpus

format. One way to do this is to transform your text into a data frame and then to
a corpus (a collection of text documents). Your R code with look similar to the
text below
library(tm)
textframe <- do.call("rbind", lapply(mytext, as.data.frame))
myCorpus <- Corpus(VectorSource(textframe$text))
Note VectorSource requires that the source is a character vector.

Tip 2: To limit the maximum number of words in the image add the max.words
to the comparision.cloud function i.e. comparision.cloud (max.word=100,…).

Tip 3: To see the list of stop words used type the following into the R console
window: stopwords("en")
Quick reference R code
install.packages("tm") # this package contains the data
install.packages("wordcloud") # this package generates the plot
require(tm)
require(wordcloud)
data(SOTU)
text <- SOTU
text <- tm_map(text, removePunctuation)
text <- tm_map(text, tolower)
text <- tm_map(text, removeNumbers)
text <- tm_map(text, function(x)removeWords(x,stopwords()))
term.matrix <- TermDocumentMatrix(text)
term.matrix <- as.matrix(term.matrix)
colnames(term.matrix) <- c("2010 Presidential Address","2011 Presidential
Address")
set.seed(1234)
par(bg="gray90")
comparison.cloud(term.matrix,max.words=300,random.order=FALSE)
NOTES

Write your notes below…

 17. State and cities with capital map

Figure 17 1 Texas cities with scale in miles

How to read the image

The image displays the scale, cities and capital city of the US state of Texas.
Major cities are denoted by white dots. Austin, the capital city is highlighted in
yellow.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("maps")# this package generates the map
library(maps)

Step 2: Clean and prepare the data

data(us.cities) # contains the data on US cities.

Step 3: Plot the chart

# set background color to brown and text to white
par(col="white",bg="brown")
# draw tan colored map of Texas
map("state", "Texas",col="tan",fill=TRUE)
# add a scale in miles to the plot
map.scale(cex=1,metric=F,relwidth=.3)
# add capital in yellow
map.cities(us.cities, country="TX", capitals=2, par(pch=23,col="yellow"))
#add Texas cities in cornsilk
map.cities(us.cities, country="TX",col="cornsilk",pch=19)
#Add title to map
title="State of Texas"
sub= "Capital and major cities"
title(title,col.main="yellow",line=2, cex.main =2)
title(sub,col.main="white",line=1, cex.main =1)

Visualization quick tips

Tip 1: The map function allows you to draw a wide variety of maps for US
states and other countries. For example map("state", "iowa") will draw an
outline of the state of Iowa. To fill in a state in a solid color use fill=TRUE.

Tip 2: Change the size of the scale by changing the cex argument in map.scale.
Values greater 1 increase the size. To use metric distances rather than imperial
set the argument metric =T.

Tip 3: You can extend the length of the scale by using the relwidth argument in
the map.scale function. The larger the value passed to the argument, the larger
the distance measured by the scale.

Quick reference R code

install.packages("maps")
library(maps)
data(us.cities)
par(col="white",bg="brown")
map("state", "Texas",col="tan",fill= TRUE)
map.scale(cex=1,metric=F,relwidth=.3)
map.cities(us.cities, country="TX", capitals=2, par(pch=23,col="yellow"))
map.cities(us.cities, country="TX",col="cornsilk",pch=19)
title="State of Texas"
sub= "Capital and major cities"
title(title,col.main="yellow",line=2, cex.main =2)
title(sub,col.main="white",line=1, cex.main =1)
NOTES

Write your notes below…

 18. Geographic pie chart map

Figure 18 1 Condiment choice for fish and chips by the British

How to read the image

The image displays the fish and chips condiment choice by the consumers in
London, England; Cardiff, Wales; Glasgow, Scotland and Belfast, Northern
Ireland. The size of the pie charts reflect the relative consumption by consumers
in each city.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("mapplots")#contains the data and draws image
require(mapplots)
Step 2: Clean and prepare the data
data(coast)# contains the data needed to draw the British Isles
xlim <- c(-10.5,3) # set longitude range on chart
ylim <- c(50,60) #set latitude range on chart
Step 3: Plot the chart
par(bg="palevioletred3")# set background color
# next we create the outline for the basic plot
basemap(xlim, ylim, main = "",xlab="",ylab="", axes = FALSE,frame.plot =
FALSE)
# now draw the map of the British isles with color set to cornsilk
draw.shape(coast, col='cornsilk')
# select the colors for the pie charts
col <- rainbow(4)
# add the pie charts
add.pie(z=c(0.7, 0.1, 0, 0.2), x=0, y=51.5,
radius=0.3,labels=NA,col=col)# London, England
add.pie(z=c(0.1, 0.1, 0.3, 0.5), x=-3.5, y=55.87,
radius=0.5,labels=NA, col=col)#Glasgow, Scotland
add.pie(z=c(0.25, 0.25, 0.25, 0.25), x=-3.5, y=52.5,
radius=0.7,labels=NA, col=col) # Cardiff,Wales
add.pie(z=c(0.2, 0.35, 0.1, 0.35), x=-6, y=54.5,
radius=0.9,labels=NA, col=col)# Belfast, Northern Ireland

# add the legend

legend.pie(2,59,labels=c('Ketchup','Brown ','Curry ','Vinegar'), radius=1, bty =
"n", col=col,
cex=0.7, label.dist=1.3)
#add a main title
title("Fish and Chips",col.main="yellow",line=2, cex.main =2)
#add a subtitle
title("Which condiment do Brits prefer?",col.main="white",line=1, cex.main
=1)
# add comments at bottom of the chart
title("*Pie size indicates frequency of consumption of Fish and
Chips",col.main="black",line=-29, cex.main =0.8)

Visualization quick tips

Tip 1: You can set a title directly in the basemap function by setting main="my
title"; If you would like to have a longitude/ latitude grid around the images set
the argument axes = TRUE and axes = FALSE,frame.plot = TRUE. Labels for
the x and y axes can be entered by xlab="x axis title",ylab="y axis title".

Tip 2: The map used the argument col <- rainbow(4) to generate four pie chart
colors using the rainbow function. A quick and easy alternatives which often
looks great is col <- terrain.colors(n), where n is the number of colors you want
to generate..

Tip 3: The add.pie function is easy and straightforward to use. The first
argument takes the values for the pie chart, the second two arguments the x and y
location of the pie chart, and the third argument the size of the pie chart. The
label.dist argument is very useful as it specifies the distance between pie chart
labels. The images used label.dist=1.3. Try experimenting with larger and
smaller values until you find a distance that is visually appealing to you.

Quick reference R code

install.packages("mapplots")
require(mapplots)
data(coast)
xlim <- c(-10.5,3)
ylim <- c(50,60)
par(bg="palevioletred3")
basemap(xlim, ylim, main = "",xlab="",ylab="", axes = FALSE,frame.plot =
FALSE)
draw.shape(coast, col='cornsilk')
col <- rainbow(4)
add.pie(z=c(0.7, 0.1, 0, 0.2), x=0, y=51.5,
radius=0.3,labels=NA,col=col)# London, England
add.pie(z=c(0.1, 0.1, 0.3, 0.5), x=-3.5, y=55.87,
radius=0.5,labels=NA, col=col)#Glasgow, Scotland
add.pie(z=c(0.25, 0.25, 0.25, 0.25), x=-3.5, y=52.5,
radius=0.7,labels=NA, col=col) # Cardiff,Wales
add.pie(z=c(0.2, 0.3, 0.1, 0.3), x=-6, y=54.5,
radius=0.9,labels=NA, col=col)# Belfast, Northern Ireland
NOTES

Write your notes below…

 19. US state income tax map

Figure 19 1 State income tax map

How to read the image

The image displays the level of state income tax by state in the United States
during 2013. The highest income tax state is California with appears shaded
purple. Zero state income tax states, such as Texas, are not colored.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("maps")#used to draw the map
install.packages("plotrix")#used to color the map
library(maps)
library(plotrix)

Step 2: Clean and prepare the data

map("usa")#load the US map
data(votes.repub) # this contains names of US States
# enter 2013 highest income tax rates by state
tax=c(5.0 ,NA ,4.5 ,7.0 ,12.3
,4.6 ,6.7 ,6.8 ,NA ,6.0 ,11.0
,7.4 ,5.0 ,3.4 ,9.0 ,4.9 ,6.0
,6.0 ,8.0 ,5.8 ,5.3 ,4.3 ,7.9
,5.0 ,6.0 ,6.9 ,6.8 ,NA ,NA
,9.0 ,4.9 ,8.8 ,7.8 ,4.0 ,5.9
,5.3 ,9.9 ,3.1 ,6.0 ,7.0 ,NA
,NA ,NA ,5.0 ,9.0 ,5.8 ,NA
,6.5 ,7.8 ,NA)

# we don’t want to waste time typing in state names, so below is a super-fast

way of getting what we want.
temp<-cbind(votes.repub,tax)
# now drop all the other columns and just leave our tax data
data <- temp[,c(32)]
# now match state names for use in the plot
state.to.map <- match.map("state", state.name, exact = F)
# store the result in a variable called x
x <- data[state.to.map]
Step 3: Plot the chart
# set the colors we will plot
state.col<-color.scale(x,c(0,300),35,50, color.spec="hcl")
par(bg="cornsilk")#set background color
# plot the map with colors
map("state",fill=TRUE,col=state.col)
#add title and subtitle
title(main="2013 State Income Tax Map",
sub="Zero income tax states not colored")
#add legend
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Low", "Low to medium", "Mid","High"), fill= state.col)

Visualization quick tips

Tip 1: For an exact match of state names set exact =TRUE in the match.map
function.

Tip 2: Playing with colors is fun. The map used the argument color.scale with
color.spec="hcl". Good alternatives are "rgb" and "hsv". The function
color.scale calculates a sequence of colors by a linear transformation of the
numeric values supplied into the ranges for the three color parameters (the map
used c(0,300),35,50). Adjust the value of each parameter to change the visual
flavor of the map.

Tip 3: You adjust the location of the legend by specifying two arguments (x and
y), these determine the horizontal and vertical location. We used a fast
alternative by specifying the location with a shortcut legend("bottomleft",…).
You can also specify "topright", "bottomright", "topleft". However, greater
precision can be obtained by specifying the x and y parameters separately.

Quick reference R code

install.packages("maps")
install.packages("plotrix")
library(maps)
library(plotrix)
map("usa")
data(votes.repub)
tax=c(5.0 ,NA ,4.5 ,7.0 ,12.3
,4.6 ,6.7 ,6.8 ,NA ,6.0 ,11.0
,7.4 ,5.0 ,3.4 ,9.0 ,4.9 ,6.0
,6.0 ,8.0 ,5.8 ,5.3 ,4.3 ,7.9
,5.0 ,6.0 ,6.9 ,6.8 ,NA ,NA
,9.0 ,4.9 ,8.8 ,7.8 ,4.0 ,5.9
,5.3 ,9.9 ,3.1 ,6.0 ,7.0 ,NA
,NA ,NA ,5.0 ,9.0 ,5.8 ,NA
,6.5 ,7.8 ,NA)
temp<-cbind(votes.repub,tax)
# we don’t want to waste time typing in state names, so this is a neat trick
# now drop all the other columns and just leave our tax data
data <- temp[,c(32)]
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]
state.col<-color.scale(x,c(0,300),35,50, color.spec="hcl")
par(bg="cornsilk")
map("state",fill=TRUE,col=state.col)
title(main="2013 State Income Tax Map",
sub="Zero income tax states not colored")
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Low", "Low to medium", "Mid","High"), fill= state.col)
NOTES

Write your notes below…

 20. US Presidential election map

Figure 20 1 Obama v Romney in battle for US presidency

How to read the image

The image displays the outcome of the 2012 US Presidential election. States
won by Obama are colored blue, states won by Romney are colored red.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("maps")#used to draw the map
install.packages("plotrix")#used to color the map
library(maps)
library(plotrix)

Step 2: Clean and prepare the data

map("usa")#load the US map
data(votes.repub) # this contains names of US States
# enter the election results note 1 equals a republican win
result=c(1 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,0 ,1 ,0
,1 ,0 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,1 ,1 ,1
,0 ,0 ,0 ,0 ,0 ,1 ,1
,0 ,1 ,0 ,0 ,0 ,1 ,1
,1 ,1 ,1 ,0 ,0 ,0 ,1
,0 ,1)
# we don’t want to waste time typing in state names, so below is a super-fast
way of getting what we want.
temp<-cbind(votes.repub,result)
# now drop all the other columns and just leave our election data
data <- temp[,c(32)]
data[data == 0] <- NA# replace zeros (Democrat) with NA’s
# match state names and store in variable x
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]

Step 3: Plot the chart

# we begin with the Republican color - red
state.col<-color.scale(x, 1,0, 0, color.spec="rgb")
par(bg="slategray2")# set background color of chart
#draw the map with Republican wins in red
map("state",fill=TRUE,col=state.col, )

# replace Republican with Democrat for next map

data[is.na(data)] <- 0.0
data[data == 1] <- NA
x <- data[state.to.map]
# set color to blue
state.col<-color.scale(x, 0,0, 1, color.spec="rgb")
#plot Democrat states in blue
map("state",fill=TRUE,col=state.col,add=TRUE, boundary = FALSE)
map("state",fill=F,add=TRUE, boundary = T, col="white")
# add title
title(main="2012 Presidential Election")
# add legend
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Obama (Democrat)", "Romney (Republican)"), fill= c("blue","red"))

Visualization quick tips

Tip 1: For an exact match of state names set exact =TRUE in the match.map
function.

Tip 2: Playing with colors is fun. The map used the argument color.scale with
color.spec="rgb" to generate the red and blue colors of the political parties.
Quick and easy alternatives are "hcl" and "hsv".

Tip 3: You adjust the location of the legend by specifying two arguments (x and
y), these determine the horizontal and vertical location. We used a fast
alternative by specifying the location with a shortcut legend("bottomleft",…).
You can also specify "topright", "bottomright", "topleft". However, greater
precision can be obtained by specifying the x and y parameters separately.

Quick reference R code

install.packages("maps")
install.packages("plotrix")
library(maps)
library(plotrix)

map("usa")
data(votes.repub)
result=c(1 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,0 ,1 ,0
,1 ,0 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,1 ,1 ,1
,0 ,0 ,0 ,0 ,0 ,1 ,1
,0 ,1 ,0 ,0 ,0 ,1 ,1
,1 ,1 ,1 ,0 ,0 ,0 ,1
,0 ,1)
temp<-cbind(votes.repub,result)
data <- temp[,c(32)]
data[data == 0] <- NA
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]
state.col<-color.scale(x, 1,0, 0, color.spec="rgb")
par(bg="slategray2")
map("state",fill=TRUE,col=state.col, )
data[is.na(data)] <- 0.0
data[data == 1] <- NA
x <- data[state.to.map]
state.col<-color.scale(x, 0,0, 1, color.spec="rgb")
map("state",fill=TRUE,col=state.col,add=TRUE, boundary = FALSE)
map("state",fill=F,add=TRUE, boundary = T, col="white")
title(main="2012 Presidential Election")
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Obama (Democrat)", "Romney (Republican)"), fill= c("blue","red"))
NOTES

Write your notes below…

 21. Population Cartogram

Figure 21 1 Governors’ political affiliation adjusted for population size

How to read the image

The image displays US Governors political affiliation by state adjusted by

population size. States held by Democratic Governors are blue and Republican
red.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("maps")#used to draw the map
library(maps)

Step 2: Clean and prepare the data

map("usa")#load the US map
data(votes.repub) # this contains names of US States
# enter the election results note 1 equals a republican win
result=c(1 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,0 ,1 ,0
,1 ,0 ,1 ,1 ,1 ,0 ,0
,0 ,0 ,0 ,1 ,1 ,1 ,1
,0 ,0 ,0 ,0 ,0 ,1 ,1
,0 ,1 ,0 ,0 ,0 ,1 ,1
,1 ,1 ,1 ,0 ,0 ,0 ,1
,0 ,1)
# we don’t want to waste time typing in state names, so below is a super-fast
way of getting what we want.
temp<-cbind(votes.repub,result)
# now drop all the other columns and just leave our election data
data <- temp[,c(32)]
data[data == 0] <- NA# replace zeros (Democrat) with NA’s
# match state names and store in variable x
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]

Step 3: Plot the chart

# we begin with the Republican color - red
state.col<-color.scale(x, 1,0, 0, color.spec="rgb")
par(bg="cornsilk")# set background color of chart
#draw the map with Republican states in red
map('state.carto',fill=TRUE,col=state.col, )
# replace Republican with Democrat for next map
data[is.na(data)] <- 0.0
data[data == 1] <- NA
x <- data[state.to.map]
# set color to blue
state.col<-color.scale(x, 0,0, 1, color.spec="rgb")
#plot Democrat states in blue
map('state.carto',fill=TRUE,col=state.col,add=TRUE, boundary = FALSE)
map('state.carto',fill=F,add=TRUE, boundary = T, col="white")
#add chart title
title(main=" Party control of Governors' offices ",
cex.main=1,col.main="black")
title( sub="As at January 2013", cex.main=0.3)
#add legend
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Democratic)", "Republican"), fill= c("blue","red"))

Visualization quick tips

Tip 1: For an exact match of state names set exact =TRUE in the match.map
function.

Tip 2: Playing with colors is fun. The map used the argument color.scale with
color.spec="rgb" to generate the red and blue colors of the political parties.
Quick and easy alternatives are "hcl" and "hsv".

Tip 3: You adjust the location of the legend by specifying two arguments (x and
y), these determine the horizontal and vertical location. We used a fast
alternative by specifying the location with a shortcut legend("bottomleft",…).
You can also specify "topright", "bottomright", "topleft". However, greater
precision can be obtained by specifying the x and y parameters separately.

Quick reference R code

install.packages("maps")
install.packages("plotrix")
library(maps)
library(plotrix)
map("usa")
data(votes.repub)
result=c(1, 1, 1 ,0 ,0 ,0
,0 ,0 ,1 ,1 ,0 ,1 ,0
,1 ,1 ,1 ,0 ,1 ,1 ,0
,0 ,1 ,0 ,1 ,0 ,0 ,1
,1 ,0 ,1 ,1 ,0 ,1 ,1
,1 ,1 ,0 ,1 ,0 ,1 ,1
,1 ,1 ,1 ,0 ,1 ,0 ,0
,1 ,1)

temp<-cbind(votes.repub,result)
data <- temp[,c(32)]
data[data == 0] <- NA
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]
state.col<-color.scale(x, 1,0, 0, color.spec="rgb")
par(bg=" cornsilk")
map('state.carto',fill=TRUE,col=state.col, )
data[is.na(data)] <- 0.0
data[data == 1] <- NA
x <- data[state.to.map]
state.col<-color.scale(x, 0,0, 1, color.spec="rgb")
map('state.carto',fill=TRUE,col=state.col,add=TRUE, boundary = FALSE)
map('state.carto',fill=F,add=TRUE, boundary = T, col="white")
title(main=" Party control of Governors' offices ",
cex.main=1,col.main="black")
title( sub="As at January 2013", cex.main=0.3)
legend(x="bottomleft", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Democratic)", "Republican"), fill= c("blue","red"))NOTES
Write your notes below…
 22. Visibility map

Figure 22 1 Governors’ political affiliation adjusted for population size

How to read the image

The image displays economic conditions across the US based on the monthly
coincident index for each of the 50 states published by the Federal Reserve
Bank of Philadelphia. The image uses a visibility based map of the US which
provides simplified state shapes with sufficient areas to allow annotations in
even the smaller states.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("maps")#used to draw the map
install.packages("plotrix")#used to color the map
library(maps)
library(plotrix)

Step 2: Clean and prepare the data

map("usa")
data(votes.repub)
# enter the data on economic conditions
conditions=c(129.23, 120.52, 181.44 ,141.03
,158.32 ,181.32 ,156.91 ,146.26
,146.20 ,166.73 ,105.82 ,199.92
,144.80 ,144.44 ,149.63 ,142.42
,141.03 ,130.47 ,135.53 ,149.65
,179.57 ,127.80 ,160.50 ,142.80
,135.51 ,160.73 ,158.23 ,182.76
,191.82 ,151.79 ,159.57 ,150.04
,162.73 ,193.45 ,140.50 ,151.11
,212.39 ,142.99 ,152.15 ,152.43
,158.02 ,153.37 ,188.18 ,199.37
,147.13 ,149.76 ,158.81 ,139.44
,163.91 ,163.22)

# we don’t want to waste time typing in state names, so this is a neat trick
temp<-cbind(votes.repub,conditions)
# now drop all the other columns and just leave our tax data
data <- temp[,c(32)]
state.to.map <- match.map("state", state.name, exact = F)
Step 3: Plot the chart
#set colors to use on the image
state.col<-color.scale(x,c(0,300),40,60, color.spec="hcl")
par(bg="cornsilk")# set background color
map("state.vbm",fill=TRUE,col=state.col)
#create title and legend
title(main="Economic Health of the Nation",
sub=" Federal Reserve Bank of Philadelphia coincident index – April 2013")
legend(x="bottomright", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Weak", "Improving", "Strong"), fill= state.col)

Visualization quick tips

Tip 1: For an exact match of state names set exact =TRUE in the match.map
function.

Tip 2: The map used the argument color.scale with color.spec="hcl". Quick and
easy alternatives are "rgb" and "hsv".

Tip 3: You adjust the location of the legend by specifying two arguments (x and
y), these determine the horizontal location and the vertical location. We used a
fast alternative by specifying the location with a shortcut legend("bottomleft",
…). You can also specify "topright", "bottomright", "topleft". However, greater
precision can be obtained by specifying the x and y parameters separately.

Quick reference R code

install.packages("maps")
install.packages("plotrix")
library(maps)
library(plotrix)
map("usa")
data(votes.repub)
conditions=c(129.23, 120.52, 181.44 ,141.03
,158.32 ,181.32 ,156.91 ,146.26
,146.20 ,166.73 ,105.82 ,199.92
,144.80 ,144.44 ,149.63 ,142.42
,141.03 ,130.47 ,135.53 ,149.65
,179.57 ,127.80 ,160.50 ,142.80
,135.51 ,160.73 ,158.23 ,182.76
,191.82 ,151.79 ,159.57 ,150.04
,162.73 ,193.45 ,140.50 ,151.11
,212.39 ,142.99 ,152.15 ,152.43
,158.02 ,153.37 ,188.18 ,199.37
,147.13 ,149.76 ,158.81 ,139.44
,163.91 ,163.22)
temp<-cbind(votes.repub,conditions)
# we don’t want to waste time typing in state names, so this is a neat trick
# now drop all the other columns and just leave our tax data
data <- temp[,c(32)]
state.to.map <- match.map("state", state.name, exact = F)
x <- data[state.to.map]
state.col<-color.scale(x,c(0,300),40,60, color.spec="hcl")
par(bg="cornsilk")
map("state.vbm",fill=TRUE,col=state.col)
title(main="Economic Health of the Nation",
sub=" Federal Reserve Bank of Philadelphia coincident index – April 2013")
legend(x="bottomright", inset=.05,
bty="n", cex=0.75, text.col ="darkred",
c("Weak", "Improving", "Strong"), fill= state.col)
NOTES

Write your notes below…

 23. Geographic map with shape observations

Figure 23 1 Latitude & longitude of Irish coast marine mammal observations

How to read the image

The image displays observations on three groups of marine mammal (whales,

dolphins and porpoise) around the Irish coast spotted by a research vessel.
Each observation is identified by a shape with each mammal group coded by
color. The longitude and latitude are captured by the y and x axis.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("mapplots")#used to draw the map
require(mapplots)

Step 2: Clean and prepare the data

data(coast)# contains the map of the Irish coastline
species=c("Fin","Humpback","Minke","Killer","Bottlenose","Common","Habour")
types of marine mammal to be monitored
# below we enter the longitude and latitude of each sighting
#data entered by mammal
#Fin whale data
fin.x=c( -5.95, -6.1, -10.48)
fin.y=c( 52.91, 52.4, 54.54)
#Humpback whale data
hump.x=c( -6.63, -6.22)
hump.y=c( 51.72, 52.13)
#Minke whale data
minke.x=c( -6.04, -10.87, -10.85, -
10.49, -10.8)
minke.y=c( 53.87, 52.3, 52.55, 54.34,
51.76)
#Killer whale data
killer.x=c( -10.61 ,-10.76)
killer.y=c( 54.5, 54.18)

#bottlenose data
bottle.x=c( -9.76 ,-9.31,-9.10)
bottle.y=c( 54.98, 55.00,54.57)
#common dolphin data
common.x=c( -10.35,-10.55)
common.y=c( 53.50,53.70)
#habour dolphin data
habour.x=c( -10.00,-10.25)
habour.y=c(53.00,52.90)

Step 3: Plot the chart

par(bg="orangered")# set background color
#draw the region used for the plot to contain the Irish coast
basemap(xlim=c(-13,-5.7), ylim=c(51.5,55.15), main = "", axes = TRUE)
draw.shape(coast, col='darkgreen') #draw the Irish coast

# Fin whale plot

draw.bubble(fin.x, fin.y, z=0.05, maxradius=0.05, pch=21, col="blue")
# Humpback whale plot
draw.bubble(hump.x, hump.y, z=0.05, maxradius=0.05, pch=22, col="blue")
# Minke whale plot
draw.bubble(minke.x, minke.y, z=0.05, maxradius=0.05, pch=23, col="blue")
# Killer whale plot
draw.bubble(killer.x, killer.y, z=0.05, maxradius=0.05, pch=24, col="blue")
# bottlenose plot
draw.bubble(bottle.x, bottle.y, z=0.05, maxradius=0.05, pch=25, col="brown")
# common plot
draw.bubble(common.x, common.y, z=0.05, maxradius=0.05, pch=21,
col="brown")
# Harbour porpoise plot
draw.bubble(habour.x, habour.y, z=0.05, maxradius=0.05, pch=22,
col="black")

#set the colors for the legend, blue for whales, brown for dolphins
colors=c("blue","blue","blue","blue","brown","brown","black")
shapes=c(21,22,23,24,25,21,22) # shapes used to plot observations
#write legend in bottom left corner of image
legend(x="bottomleft", inset=.05, bty="0", cex=0.75, text.col ="white",
pch=shapes,col=colors,legend=species, title="Key",title.col="black")
#add title to image
title(main="Marine Mammal observations around the Irish Coast",
sub=" 30th May – 15th April 2013: Irish Research Vessel")

Visualization quick tips

Tip 1: The draw.bubble function was used to draw the observed marine
mammal sighting on the image. The size and radius of each point can be changed
by specifying alternative values for the arguments z and maxradius respectively.

Tip 2: To turn off the x and y axis use the argument axes = FALSE in the
basemap function.

Tip 3: You adjust the location of the legend by specifying two arguments (x and
y), these determine the horizontal location and the vertical location. We used a
fast alternative by specifying the location with a shortcut legend("bottomleft",
…). You can also specify "topright", "bottomright", "topleft". However, greater
precision can be obtained by specifying the x and y parameters separately.

Quick reference R code

install.packages("mapplots")
require(mapplots)
data(coast)
species=c("Fin","Humpback","Minke","Killer","Bottlenose","Common","Habour")

#Fin whale data

fin.x=c( -5.95, -6.1, -10.48)
fin.y=c( 52.91, 52.4, 54.54)
#Humpback whale data and plot
hump.x=c( -6.63, -6.22)
hump.y=c( 51.72, 52.13)
#Minke whale data and plot
minke.x=c( -6.04, -10.87, -10.85, -
10.49, -10.8)
minke.y=c( 53.87, 52.3, 52.55, 54.34,
51.76)
#Killer whale data and plot
killer.x=c( -10.61 ,-10.76)
killer.y=c( 54.5, 54.18)

#bottlenose data and plot

bottle.x=c( -9.76 ,-9.31,-9.10)
bottle.y=c( 54.98, 55.00,54.57)
#common dolphin data and plot
common.x=c( -10.35,-10.55)
common.y=c( 53.50,53.70)
#habour dolphin data and plot
habour.x=c( -10.00,-10.25)
habour.y=c(53.00,52.90)

par(bg="orangered")
basemap(xlim=c(-13,-5.7), ylim=c(51.5,55.15), main = "", axes = TRUE)
draw.shape(coast, col='darkgreen')

# Fin whale plot

draw.bubble(fin.x, fin.y, z=0.05, maxradius=0.05, pch=21, col="blue")
# Humpback whale plot
draw.bubble(hump.x, hump.y, z=0.05, maxradius=0.05, pch=22, col="blue")
# Minke whale plot
draw.bubble(minke.x, minke.y, z=0.05, maxradius=0.05, pch=23, col="blue")
# Killer whale plot
draw.bubble(killer.x, killer.y, z=0.05, maxradius=0.05, pch=24, col="blue")
# bottlenose plot
draw.bubble(bottle.x, bottle.y, z=0.05, maxradius=0.05, pch=25, col="brown")
# common plot
draw.bubble(common.x, common.y, z=0.05, maxradius=0.05, pch=21,
col="brown")
# Harbour porpoise plot
draw.bubble(habour.x, habour.y, z=0.05, maxradius=0.05, pch=22,
col="black")

colors=c("blue","blue","blue","blue","brown","brown","black")#legend colors
shapes=c(21,22,23,24,25,21,22)
legend(x="bottomleft", inset=.05, bty="0", cex=0.75, text.col ="white",
pch=shapes,col=colors,legend=species,title="Key",title.col="black")
title(main="Marine Mammal observations around the Irish Coast",
sub=" 30th May – 15th April 2013: Irish Research Vessel")
NOTES

Write your notes below…

 24. Two variable 3D scatter plot

Figure 24 1 Bacteria count, weight over time of two samples of raw milk

How to read the image

The image displays the relationship between bacteria count, weight (mass) and
time of two samples of raw milk. Sample one is colored green and sample two
colored red.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d")# this package generates the plot
require("scatterplot3d")

Step 2: Clean and prepare the data

# Time in minutes
time=c(25,37,49,61,73,85,97,109,121,133,145,157,169,181,193,205,217)
#Sample 1 bacteria count and weight reading
count.1=c(0,7,9,12,7,30,8,12,33,16,16,30,29,19,28,29,11)
mass.1=c(220,110,210,340,330,560,635,80,605,150,688,422,1511,642,681,1961,1645)
#Sample 2 bacteria count and weight reading
count.2=c(6,13,15,18,13,36,14,18,39,22,22,36,35,25,34,35,17)
mass.2=c(239,197,213,481,466,591,695,140,746,203,782,494,1562,757,796,2055,1739

Step 3: Plot the chart

# set color of plotted points to green and background color to cornsilk
par(col="green",bg="cornsilk")
#plot the first sample points
scatterplot3d(ylim=c(0,1600),zlim=c(0,2000) , count.1, time, mass.1, angle=22,
pch=20, cex.symbols=1,xlab="", ylab="",
zlab="",grid=FALSE,axis=FALSE,cex.lab=1, cex.axis=1)
# set color of plotted points to red and overlay scatterplot over first sample
points
par(new=TRUE,col="red" )

scatterplot3d(main=" Two samples of raw milk",ylim=c(0,1600),

zlim=c(0,2000), count.2, time, mass.2, angle=22, col.axis="grey5"
,col.grid="tan",xlab=" Bacteria Count", ylab="Mass/ weight" ,zlab="Time
(minutes)", pch=20, cex.symbols=1, box=FALSE, col.lab="brown",cex.lab=1,
cex.axis=1)

Visualization quick tips

Tip 1: You can set the limit range of the x,y and z axis by using the arguments
xlim, ylim and zlim respectively. So zlim=c(0,2000) sets the range of the z axis
between 0 and 2000.

Tip 2: To set the size of the labels, axis and symbols (plotted points) use the
argument cex.lab, cex.axis and cex.symbols respectively. Try experimenting
with values greater than 1 for much larger labels, symbols or x, y and z axis,
values less than 1 produce smaller labels, symbols and x,y,z axis.
Tip 3: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.
Quick reference R code
install.packages("scatterplot3d")
require("scatterplot3d")
time=c(25,37,49,61,73,85,97,109,121,133,145,157,169,181,193,205,217)
count.1=c(0,7,9,12,7,30,8,12,33,16,16,30,29,19,28,29,11)
mass.1=c(220,110,210,340,330,560,635,80,605,150,688,422,1511,642,681,1961,1645)
count.2=c(6,13,15,18,13,36,14,18,39,22,22,36,35,25,34,35,17)
mass.2=c(239,197,213,481,466,591,695,140,746,203,782,494,1562,757,796,2055,1739
par(col="green",bg="cornsilk")
scatterplot3d(ylim=c(0,1600),zlim=c(0,2000) , count.1, time, mass.1, angle=22,
pch=20, cex.symbols=1,xlab="",ylab="", zlab="",grid=F,axis=F,cex.lab=1,
cex.axis=1)
par(new=TRUE,col="red" )
scatterplot3d(main=" Two samples of raw
milk",ylim=c(0,1600),zlim=c(0,2000), count.2, time, mass.2, angle=22,
col.axis="grey5",col.grid="tan",xlab=" Bacteria Count",ylab="Mass/ weight",
zlab="Time (minutes)", pch=20, cex.symbols=1,box=FALSE,
col.lab="brown",cex.lab=1, cex.axis=0.3)
legend("topright", inset=.05, bty="o", cex=1,
title="Key", c("Sample 1", "Sample 2"), pch=20, fill=c("green",
"red"),text.col="brown")
NOTES

Write your notes below…

 25. 3D scatter plot

Figure 25 1 Bacteria count, weight over time of a raw milk sample

How to read the image

The image displays the relationship between bacteria count, weight (mass) and
time from a sample of raw milk. Sample points are red colored dots.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d")# this package generates the plot
require("scatterplot3d")

Step 2: Clean and prepare the data

# Time in minutes
time=c(25,37,49,61,73,85,97,109,121,133,145,157,169,181,193,205,217)
# bacteria count and weight reading
count=c(0,7,9,12,7,30,8,12,33,16,16,30,29,19,28,29,11)
mass=c(220,110,210,340,330,560,635,80,605,150,688,422,1511,642,681,1961,1645)

Step 3: Plot the chart

# set color of plotted points and background color
par(col="red",bg="white")
# plot 3D scatter plot
scatterplot3d(main="Raw Milk Sample",count, time, mass, angle=22,
col.axis="lightblue",col.grid="lightblue",xlab=" Bacteria Count",ylab="Mass/
weight", zlab="Time (minutes)", pch=20, cex.symbols=1, box=TRUE,
col.lab="blue",cex.lab=1, cex.axis=1)

Visualization quick tips

Tip 1: You can set the limit range of the x,y and z axis by using the arguments
xlim, ylim and zlim respectively. So zlim=c(0,2000) sets the range of the z axis
between 0 and 2000.

Tip 2: To set the size of the labels, axis and symbols (plotted points) use the
argument cex.lab, cex.axis and cex.symbols respectively. Try experimenting
with values greater than 1 for much larger labels, symbols or x, y and z axis,
values less than 1 produce smaller labels, symbols and x,y,z axis.

Tip 3: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.

Quick reference R code

install.packages("scatterplot3d")
require("scatterplot3d")
time=c(25,37,49,61,73,85,97,109,121,133,145,157,169,181,193,205,217)
count=c(0,7,9,12,7,30,8,12,33,16,16,30,29,19,28,29,11)
mass=c(220,110,210,340,330,560,635,80,605,150,688,422,1511,642,681,1961,1645)
par(col="red",bg="white")
scatterplot3d(main=" Raw Milk Sample",count, time, mass, angle=22,
col.axis="lightblue",col.grid="lightblue",xlab=" Bacteria Count",ylab="Mass/
weight",zlab="Time (minutes)", pch=20, cex.symbols=1,box=TRUE,
col.lab="blue",cex.lab=1, cex.axis=1)
NOTES

Write your notes below…

 26. 3D cluster scatter plot

Figure 26 1 Clusters from three samples of normal random variables

How to read the image

The image displays the values from three distinct samples from the normal
distribution. Each sample has a different mean, although all three samples have
the same variance. Clusters are color coded with increasing mean as you move
from the green cluster through the blue cluster, with the orange cluster having the
highest mean value.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d"))# this package generates the plot
require("scatterplot3d")
Step 2: Clean and prepare the data
set.seed(123)# allows image to be replicated
#simulate 300 observations for group1
sample1<-matrix(rnorm(300,0,1),100,3)
group1<- rep(1,nrow(sample1) )
sample1<-cbind(sample1,group1)
#simulate 300 observations for group2
sample2<-matrix(rnorm(300,2,1),100,3)
group2<- rep(2,nrow(sample2) )
sample2<-cbind(sample2,group2)
#simulate 300 observations for group3
sample3<-matrix(rnorm(300,4,1),100,3)
group3<- rep(3,nrow(sample3) )
sample3<-cbind(sample3,group3)
#combine the three samples together and add column names
data=rbind(sample1,sample2,sample3)
data<-data.frame(data)
colnames(data)=c("x","y","z","Group")

Step 3: Plot the chart

#define the colors to be plotted for each group
data$pcolor[data$Group==1] <- "green"
data$pcolor[data$Group ==2] <- "blue"
data$pcolor[data$Group ==3] <- "orange"
par(bg="ghostwhite")#set background color of image
#plot the image and add a legend
with(data, {
image<-scatterplot3d(x,y,z, type="p", pch=19,
color=pcolor,box=TRUE,xlab="value of x",zlab="value of y",ylab="value of
z",angle=45,main="Random Normal Clusters", col.main="darkblue",
col.grid="grey")
legend("bottomleft", inset=.05,
bty="n", cex=1, text.col ="darkred", title="", c("Cluster 1", "Cluster 2",
"Cluster 3"), fill=c("green", "blue", "orange"))
})

Visualization quick tips

Tip 1: The argument type determines the type of plot. Use type = "p" for points,
type = "l" for lines and type = "h" for vertical lines.
Tip 2: Change the angle of the image by using the angle argument. Values in the
range of 20 to 45 generally work well. Notice that angle=0 results in a 2-D plot.

Tip 3: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.
Tip 4: use the $pcolor argument to set the color of each group of bars.

Quick reference R code

install.packages("scatterplot3d")
require(scatterplot3d)
set.seed(123)
sample1<-matrix(rnorm(300,0,1),100,3)
group1<- rep(1,nrow(sample1) )
sample1<-cbind(sample1,group1)

sample2<-matrix(rnorm(300,2,1),100,3)
group2<- rep(2,nrow(sample2) )
sample2<-cbind(sample2,group2)

sample3<-matrix(rnorm(300,4,1),100,3)
group3<- rep(3,nrow(sample3) )
sample3<-cbind(sample3,group3)

data=rbind(sample1,sample2,sample3)
data<-data.frame(data)
colnames(data)=c("x","y","z","Group")

data$pcolor[data$Group==1] <- "green"

data$pcolor[data$Group ==2] <- "blue"
data$pcolor[data$Group ==3] <- "orange"
par(bg="ghostwhite")
with(data, {
image<-scatterplot3d(x,y,z, type="p", pch=19,
color=pcolor,box=TRUE,xlab="value of x",zlab="value of y",ylab="value of
z",angle=45,main="Random Normal Clusters", col.main="darkblue",
col.grid="grey")
legend("bottomleft", inset=.05,
bty="n", cex=1, text.col ="darkred", title="",
c("Cluster 1", "Cluster 2", "Cluster 3"), fill=c("green", "blue", "orange"))
})
NOTES

Write your notes below…

 27. Vertical line with points 3D scatter plot

Figure 27 1 Path of US interest rates and inflation during 1990

How to read the image

The figure displays the path of inflation and interest rates in the US during 1990.
Each sample point is represented by a vertical dashed line and an open circle.
The solid purple line captures the trajectory over time of interest rates and
inflation.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d"))# this package generates the plot
install.packages("Ecdat"))# this package contains the data
require(Ecdat)
require("scatterplot3d")

Step 2: Clean and prepare the data

# use the Mishkin dataset contained in package Ecdat
data(Mishkin)
#create label for image
label=c("Dec","Feb","Apr","Jun","Aug","Oct")
# define axis variables
inflation=Mishkin[480:491,2]
rates=Mishkin[480:491,4]
time<-(1:12)# Time in months

Step 3: Plot the chart

# set color of plotted points to blue and background color to azure
par(bg="burlywood1",col="blue")
scatterplot3d(time,inflation,rates,type="l",box=F, tick.marks=T,
label.tick.marks=T, x.ticklabs=label,angle=25,xlab="Date",ylab="Interest Rate
(%)", zlab="Inflation Rate")

par(new=TRUE,col="blue" )

scatterplot3d(time,inflation,rates,type="h",box=F, tick.marks=T,
label.tick.marks=T, x.ticklabs=label,angle=25,xlab="Date",ylab="Interest Rate
(%)", zlab="Inflation Rate", lty.hplot="dotted",col.grid="red",main= "US
Inflation and Interest Rates - 1990",col.main="firebrick4") Visualization quick
tips

Tip 1: The argument type determines the type of plot. Use type = "p" for points,
type = "l" for lines and type = "h" for vertical lines.

Tip 2: The arguments tick.marks and label.tick.marks contain a logical value

indicating whether tick marks should be drawn on the plot and whether tick
marks should be labeled on the plot respectively.

Tip 3: The argument lty.hplot="dotted" determines the nature of the vertical line.
In this case dotted. Other values the argument can take include "blank", "solid",
"dashed", "dotdash", "longdash", or "twodash".

Tip 4: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.

Quick reference R code

install.packages("scatterplot3d")
install.packages("Ecdat")
require(Ecdat)
require("scatterplot3d")
data(Mishkin )
label=c("Dec","Feb","Apr","Jun","Aug","Oct")
inflation=Mishkin[480:491,2]
rates=Mishkin[480:491,4]
time<-(1:12)
par(bg="burlywood1",col="blue")
scatterplot3d(time,inflation,rates,type="l",box=F, tick.marks=T,
label.tick.marks=T, x.ticklabs=label,angle=25,xlab="Date",ylab="Interest Rate
(%)", zlab="Inflation Rate")
par(new=TRUE,col="blue" )
scatterplot3d(time,inflation,rates,type="h",box=F, tick.marks=T,
label.tick.marks=T, x.ticklabs=label,angle=25,xlab="Date",ylab="Interest Rate
(%)", zlab="Inflation Rate", lty.hplot="dotted",col.grid="red",main= "US
Inflation and Interest Rates - 1990",col.main="firebrick4")
NOTES

Write your notes below…

 28. Vertical labeled 3D scatter plot

Figure 28 1 Savings characteristics of six nations

How to read the image

The image displays the savings rate, growth in disposable income and
population over 75 for six nations averaged over the decade 1960–1970. Each
nation is represented by a red vertical dashed line with an open circle.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d"))# this package generates the plot
require("scatterplot3d")
Step 2: Clean and prepare the data
require(datasets)# contains the data frame used
attach(LifeCycleSavings)# contains the data
# we select six nations from the LifeCycleSavings data frame
nations<-rbind( LifeCycleSavings[8,], LifeCycleSavings[11,],
LifeCycleSavings[14,], LifeCycleSavings[15,], LifeCycleSavings[20,],
LifeCycleSavings[23,])
# axis variables to be used in plot
x=nations[,1]
y=nations[,3]
z=nations[,5]

Step 3: Plot the chart

# set background color and color of plotted points
par(bg="ghostwhite",col="firebrick1")
# details of plot
image <-scatterplot3d(x,y,z,box=FALSE,type="h",angle=45, xlab="Savings
Rate", ylab="Population over 75 (%)",zlab="Growth in disposable income
(%)",main="Saving Characteristics",sub="1960-1970",col.main="darkblue",
col.grid="grey",color="firebrick1")
# get image coordinates to enable plotting to country names
image.coords <- image$xyz.convert(x,y,z)
# add nation names to the vertical line
text(image.coords$x, image.coords$y, labels= row.names (nations) , pos=3,
cex=1)

Visualization quick tips

Tip 1: The argument type determines the type of plot. Use type = "p" for points,
type = "l" for lines and type = "h" for vertical lines.

Tip 2: Change the angle of the image by using the angle argument. Values in the
range of 20 to 45 generally work well. Notice that angle=0 results in a 2-D plot.

Tip 3: Add the argument lty.hplot="dotted" to change the nature of the vertical
line. Argument can take the form "blank", "solid", "dashed", "dotted", "dotdash",
"longdash", or "twodash".

Tip 4: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.
Quick reference R code
install.packages("scatterplot3d")
require(scatterplot3d)
require(datasets)
attach(LifeCycleSavings)
nations<-rbind( LifeCycleSavings[8,], LifeCycleSavings[11,],
LifeCycleSavings[14,], LifeCycleSavings[15,], LifeCycleSavings[20,],
LifeCycleSavings[23,])
x=nations[,1]
y=nations[,3]
z=nations[,5]
par(bg="ghostwhite",col="firebrick1")
image <-scatterplot3d(x,y,z,box=FALSE,type="h",angle=45, xlab="Savings
Rate",ylab="Population over 75 (%)",zlab="Growth in disposable income
(%)",main="Saving Characteristics",sub="1960-1970",col.main="darkblue",
col.grid="grey",color="firebrick1")
image.coords <- image$xyz.convert(x,y,z)
text(image.coords$x, image.coords$y, labels=row.names(nations),
pos=3, cex=1)
NOTES

Write your notes below…

 29. 3D grid color bar plot

Figure 29 1 Asset class return and risk expectations

How to read the image

The image displays the expected asset class returns and risk for stocks (large
cap, small cap, international), bonds (government, corporate, high yield) and
real assets (real estate, commodities, gold) alongside the current portfolio asset
allocation. Each asset class is color coded, stocks (green), bonds(blue) and real
assets (orange). The name of each asset appear at the top of the colored bar.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("scatterplot3d"))# this package generates the plot
require("scatterplot3d")
Step 2: Clean and prepare the data
# create a variable for each asset. It contains expected return, expected risk,
current portfolio asset allocation and asset group (1 for stocks, 2 for bonds and
3 for real assets)
large.cap=c(8, 16, 20,1)
small.cap=c(9, 15, 20,1)
international=c(10, 17, 15,1)
Corp.bonds=c( 7, 8, 5,2)
Gov.bonds=c(2, 5, 4,2)
High.yield=c( 8, 12, 5,2)
Real.estate=c( 10, 18, 10,3)
Commodities=c( 8, 22, 10,3)
Gold=c( 2, 25, 10,3)
data=rbind(large.cap, small.cap, international, Corp.bonds, Gov.bonds,
High.yield, Real.estate, Commodities, Gold)
Add row names and create a data frame then add column names
rownames(data)=c("Large","Small","Int","Corp","Gov","High","Real","Comd","Gold")
data<-data.frame(data)
colnames(data)=c("ER","Vol","%","Group")

data<-data.frame(data)
# data used for axis in plot
x<-data[,2]
y<-data[,3]
z<-data[,1]

Step 3: Plot the chart

#define colors for each group to be plotted on image
data$pcolor[data$Group==1] <- "green"
data$pcolor[data$Group ==2] <- "blue"
data$pcolor[data$Group ==3] <- "orange"
#set background color
par(bg="ghostwhite")
# Create plot
with(data, {
image<-scatterplot3d(x,y,z, type="h", lwd=10, pch=" ", color=pcolor,zlim =
c(0, 12),box=FALSE,xlab="Risk (vol)",zlab="Expected Return",ylab="Current
Allocation",angle=22,main="Asset Class Expectation", col.main="darkblue",
col.grid="grey")
#overlay text names of each asset
image.coords <- image$xyz.convert(x,y,z)
text(image.coords$x, image.coords$y,
labels=row.names(data),
pos=3, cex=1, col="darkred")
#add legend
legend("topleft", inset=.05,
bty="n", cex=1, text.col ="darkred",
title="Asset Class",
c("Stocks", "Bonds", "Real Assets"), fill=c("green", "blue", "orange"))
})

Visualization quick tips

Tip 1: The argument type determines the type of plot. Use type = "p" for points,
type = "l" for lines and type = "h" for vertical lines.

Tip 2: Change the angle of the image by using the angle argument. Values in the
range of 20 to 45 generally work well. Notice that angle=0 results in a 2-D plot.

Tip 3: Add the argument lty.hplot="dotted" to change the nature of the vertical
line. Argument can take the form "blank", "solid", "dashed", "dotted", "dotdash",
"longdash", or "twodash".

Tip 4: Set the color of the grid and axes by using the arguments col.axis and
col.grid. You can turn the axis and grid off by setting the arguments axis=FALSE
and grid =FALSE.
Tip 5: use the $pcolor argument to set the color of each group of bars.

Quick reference R code

install.packages("scatterplot3d")
require(scatterplot3d)

large.cap=c(8, 16, 20,1)

small.cap=c(9, 15, 20,1)
international=c(10, 17, 15,1)
Corp.bonds=c( 7, 8, 5,2)
Gov.bonds=c(2, 5, 4,2)
High.yield=c( 8, 12, 5,2)
Real.estate=c( 10, 18, 10,3)
Commodities=c( 8, 22, 10,3)
Gold=c( 2, 25, 10,3)

data=rbind(large.cap, small.cap, international, Corp.bonds, Gov.bonds,

High.yield, Real.estate, Commodities, Gold)
rownames(data)=c("Large","Small","Int","Corp","Gov","High","Real","Comd","Gold")
data<-data.frame(data)
colnames(data)=c("ER","Vol","%","Group")

data<-data.frame(data)
x<-data[,2]
y<-data[,3]
z<-data[,1]
data$pcolor[data$Group==1] <- "green"
data$pcolor[data$Group ==2] <- "blue"
data$pcolor[data$Group ==3] <- "orange"

par(bg="ghostwhite")

with(data, {
image<-scatterplot3d(x,y,z, type="h", lwd=10, pch=" ", color=pcolor,zlim =
c(0, 12),box=FALSE,xlab="Risk (vol)",zlab="Expected Return",ylab="Current
Allocation",angle=22,main="Asset Class Expectation", col.main="darkblue",
col.grid="grey")

image.coords <- image$xyz.convert(x,y,z)

text(image.coords$x, image.coords$y,
 labels=row.names(data),
pos=3, cex=1, col="darkred")

legend("topleft", inset=.05,
bty="n", cex=1, text.col ="darkred",

title="Asset Class",
c("Stocks", "Bonds", "Real Assets"), fill=c("green", "blue", "orange"))
})
NOTES

Write your notes below…

 30. Circle plot

Figure 30 1 Simulated flight direction of three individual files

How to read the chart

The chart displays the simulated flight directions of three flies from a point of
rest. The colored dots represent the flight orientation of each fly (red for fly 1,
blue for fly 2 and green for fly 3).

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package generates the plot
require(circular)

Step 2: Clean and prepare the data

set.seed(1234)# set so results can be replicated
#the code below simulates flight orientation from von mises distribution. We
observe 50 flight paths per fly.
fly.1 <- rvonmises(n=50, mu=circular(0), kappa=3)
fly.2 <- rvonmises(n=50, mu=circular(pi/3), kappa=3)
fly.3 <- rvonmises(n=50, mu=circular(pi/1.5), kappa=3)

Step 3: Plot the chart

res <- plot(fly.1, stack=TRUE, col="darkred") # plot fly 1

#below adds data to chart for fly 2 and fly 3

points(fly.2, plot.info=res, col="darkblue", stack=TRUE)

points(fly.3, plot.info=res, col="darkgreen", stack=TRUE)

Visualization quick tips

Tip 1: Data should be in a circular format. This means your data needs to be
converted to circular class. To do this try the statement circular (your_data) to
coerce data into a circular format.

Tip 2: You can change the color of the points by changing the col argument.

Tip 3: Set stack=FALSE to plot the points around the outer edge of the circle.

Tip 4: To remove the numbers around the edge of the circle add the argument
axes=FALSE to the plot function. To add tick marks to the edge of the circle add
the argument ticks =TRUE to the plot function.

Quick reference R code

install.packages("circular")
require(circular)
data<-circular(wind , zero = pi/2)
par(bg="lightyellow")
rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="darkred",tol=0,prop=1.5)
points(data,stack=TRUE,col="darkgreen",pch=20)
symbols(0, 0, circle = 0.2, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.4, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.6, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.8, inches = FALSE, add = TRUE, fg = 'black')
NOTES

Write your notes below…

 31. Circular arrows split plot

Figure 31 1 Direction of three groups of long legged desert ants

How to read the chart

A arrows chart for circular data shows the orientation of each individual data
point by use of an arrow. The chart displays the orientation of three sets of long-
legged desert ants after one eye on each ant was 'trained' to learn the ant's home
direction, then covered and the other eye uncovered. The charts are orientated
so that zero degrees appears at the top.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
# The data frame fisherB10 contains observations on each of the ant groups. It is
converted to an object of class circular and orientated so that zero equals pi/2.
This will ensure zero appears at the top of the resultant arrow plots.

Step 3: Plot the chart

par(bg="cornsilk")# set the background color to cornsilk
# create a 2 by 2 area, each of the four charts is plotted on one area
par(mfrow=c(2,2))
#plot the first chart with arrows
plot(ant.1,col="darkred",main="Ant group 1",axes=FALSE)
arrows.circular (ant.1,col="darkred")
#plot the second chart with arrows
plot(ant.2,col="darkgreen",main="Ant group 2",axes=FALSE)
arrows.circular (ant.2,col="darkgreen")
#plot the third chart with arrows
plot(ant.3,col="darkblue",main="Ant group 3",axes=FALSE)
arrows.circular (ant.3,col="darkblue")
#plot all the data with arrows
plot(ant.1,col="darkred",main="All three groups",axes=FALSE)
arrows.circular (ant.1,col="darkred")
par(new=TRUE) # allows you plot on an existing image
plot(ant.2,col="darkgreen", axes=FALSE)
arrows.circular (ant.2,col="darkgreen")
par(new=TRUE) # allows you plot on an existing image
plot(ant.3,col="darkblue",axes=FALSE)
arrows.circular (ant.3,col="darkblue")

Visualization quick tips

Tip 1: The arrows.circular function requires data in a circular format. This

means the data needs to be converted to circular class. When using your own
data or other datasets you can use the statement circular (your_data) to coerce
data into a circular format.
Tip 2: You can include axes on each chart by setting the argument axes=TRUE
in the plot function.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Tip 4: To add tick marks around the edge of the circle add the argument ticks
=TRUE to the plot function.

Quick reference R code

install.packages("circular")
require(circular)
ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
par(bg="cornsilk")
par(mfrow=c(2,2))
plot(ant.1,col="darkred",main="Ant group 1",axes=FALSE)
arrows.circular (ant.1,col="darkred")
plot(ant.2,col="darkgreen",main="Ant group 2",axes=FALSE)
arrows.circular (ant.2,col="darkgreen")
plot(ant.3,col="darkblue",main="Ant group 3",axes=FALSE)
arrows.circular (ant.3,col="darkblue")
plot(ant.1,col="darkred",main="All three groups",axes=FALSE)
arrows.circular (ant.1,col="darkred")
par(new=TRUE)
plot(ant.2,col="darkgreen", axes=FALSE)
arrows.circular (ant.2,col="darkgreen")
par(new=TRUE)
plot(ant.3,col="darkblue",axes=FALSE)
arrows.circular (ant.3,col="darkblue")

NOTES
Write your notes below…
 32. Rose diagram

Figure 32 1 Wind direction measurements at Col d al Rosa

How to read the chart

A rose diagram is essentially a histogram for circular data. The chart displays
wind direction measurement taken at 15 minute intervals at Col de la Roa in the
Italian Alps from January 29, 2001 to March 31, 2001. The dark red bars in the
center of the diagram are similar to the bars in a histogram. The image is
orientated so that zero degrees appears at the top of the image.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

data<-circular(wind , zero = pi/2)
# The data frame wind contains the observations. It is converted to an object of
class circular and orientated so that zero equals pi/2. This will ensure zero
appears at the top of the resultant plot.

Step 3: Plot the chart

par(bg="lightyellow")# set the background color to light yellow

rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="darkred",tol=0,prop=1.5)

Visualization quick tips

Tip 1: The rose.diag function requires data in a circular format. This means the
data needs to be converted to circular class. When using your own data or other
datasets you can use the statement circular (your_data) to coerce data into a
circular format.

Ti p 2: You can change the number of bins by using the arguments bins in
rose.diag. The argument takes positive values.

Tip 3: Change the argument col in rose.diag to choose different bin colors.

Tip 4: To remove the numbers around the edge of the circle use axes=FALSE.
To remove the tick marks from the edge of the circle use ticks =FALSE.

Quick reference R code

install.packages("circular")
require(circular)
data<-circular(wind , zero = pi/2)
par(bg="lightyellow")
rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="darkred",tol=0,prop=1.5)
points(data,stack=TRUE,col="darkgreen",pch=20)
symbols(0, 0, circle = 0.2, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.4, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.6, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.8, inches = FALSE, add = TRUE, fg = 'black')
NOTES

Write your notes below…

 33. Stacked rose diagram

Figure 33 1 Stacked wind direction measurements at Col d al Rosa

How to read the chart

A stacked rose diagram is essentially a histogram for circular data with

observations plotted at the edge. The chart displays wind direction measurement
taken at 15 minute intervals at Col de la Roa in the Italian Alps from January 29,
2001 to March 31, 2001. The dark red bars in the center of the diagram are
similar to the bars in a histogram. The green dots around the edge of the outer
circle are the stacked individual observations. The diagram is orientated so that
zero degrees appears at the top of the image.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

data<-circular(wind , zero = pi/2)
# The data frame wind contains the observations. It is converted to an object of
class circular and orientated so that zero equals pi/2. This will ensure zero
appears at the top of the resultant plot.

Step 3: Plot the chart

par(bg="lightyellow")# set the background color to light yellow
rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="darkred",tol=0,prop=1.5) # plot the rose diagram
points(data,stack=TRUE,col="darkgreen",pch=20)# add individual
observations

Visualization quick tips

Tip 1: The rose.diag function requires data in a circular format. This means the
data needs to be converted to circular class. When using your own data or other
datasets you can use the statement circular (your_data) to coerce data into a
circular format.

Ti p 2: You can change the number of bins by using the arguments bins in
rose.diag. The argument takes positive values.

Tip 3: Change the argument col in rose.diag to choose different bin colors.

Tip 4: In the points function set Stack =FALSE if you want individual
observations plotted around the edge of the outer circle.

Quick reference R code

install.packages("circular")
require(circular)
data<-circular(wind , zero = pi/2)
par(bg="lightyellow")
rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="darkred",tol=0,prop=1.5)
points(data,stack=TRUE,col="darkgreen",pch=20)
 34. Radial rose diagram

Figure 34 1 Wind direction measurements at Col d al Rosa

How to read the chart

A radial rose diagram is essentially a histogram for circular data with radial
lines indicating the frequency of observations. The chart displays wind direction
measurement taken at 15 minute intervals at Col de la Roa in the Italian Alps
from January 29, 2001 to March 31, 2001. The gold petals in the center of the
diagram are similar to the bars in a histogram. The green dots around the edge of
the outer circle are the stacked individual observations. The diagram is
orientated so that zero degrees appears at the top of the image.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

data<-circular(wind , zero = pi/2)
# The data frame wind contains the observations. It is converted to an object of
class circular and orientated so that zero equals pi/2. This will ensure zero
appears at the top of the resultant plot.

Step 3: Plot the chart

par(bg="lightblue") # set background color to light blue
rose.diag(data, bins=18,axes=TRUE,ticks=TRUE,border="black",
col="gold",tol=0,prop=1.5) # function to plot chart
# line below adds individual observations
points(data,stack=TRUE,col="darkred",pch=20)
# lines below add radials
symbols(0, 0, circle = 0.2, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.4, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.6, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.8, inches = FALSE, add = TRUE, fg = 'black')

Visualization quick tips

Tip 1: The rose.diag function requires data in a circular format. This means the
data needs to be converted to circular class. When using your own data or other
datasets you can use the statement circular (your_data) to coerce data into a
circular format.

Ti p 2: You can change the number of bins by using the arguments bins in
rose.diag. The argument takes positive values.

Tip 3: Change the argument col in rose.diag to choose different bin colors. In the
points function set Stack =FALSE if you want individual observations plotted
around the edge of the outer circle.

Tip 4: You can change the color of the radial lines by setting the argument
fg='your chosen color' in the symbols function.
Quick reference R code

install.packages("circular") # this package contains the data and generates the

plot
require(circular)
data<-circular(wind , zero = pi/2)

par(bg="lightblue")
rose.diag(data,
bins=18,axes=TRUE,ticks=TRUE,border="black",col="gold",tol=0,prop=1.5)
#
points(data,stack=TRUE,col="darkred",pch=20)
symbols(0, 0, circle = 0.2, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.4, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.6, inches = FALSE, add = TRUE, fg = 'black')
symbols(0, 0, circle = 0.8, inches = FALSE, add = TRUE, fg = 'black')
NOTES

Write your notes below…

 35. Stacked multivariable rose diagram

Figure 35 1 Simulated forage orientation of three beavers

How to read the chart

A multivariable rose diagram is essentially a grouped histogram for circular

data. The chart displays simulated forage direction of three beavers. The dark
red petals in the center of the diagram are similar to the bars in a histogram and
represent beaver 1. Beaver 2 is represented by the blue petals, and beaver 3 by
the orange petals. The dots around the edge of the outer circle are the color
coded stacked observations for each beaver.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)
# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

set.seed(1234)# allows results to be replicated
#below the simulated orientation of the beavers is simulated using a von mises
distribution.
beaver.1 <- rvonmises(n=50, mu=circular(0), kappa=3)
beaver.2 <- rvonmises(n=50, mu=circular(pi), kappa=3)
beaver.3 <- rvonmises(n=50, mu=circular(pi/2), kappa=3)

Step 3: Plot the chart

par(bg="cornsilk")# set background color
# print rose diagram for beaver 1
rose.diag(beaver.1, bins=18,axes=TRUE,ticks=TRUE,
border="black",col="darkred",tol=0,prop=1.5)
par(new=TRUE) # allow overlay of next image
# print rose diagram for beaver 2
rose.diag(beaver.2, bins=18,axes=TRUE,ticks=TRUE
,border="black",col="darkblue",tol=0,prop=1.5) #
par(new=T) # allow overlay of next image

# print rose diagram for beaver 3

rose.diag(beaver.3, bins=18,axes=TRUE,ticks=TRUE,
border="black",col="darkorange",tol=0,prop=1.5) #
#plot stacked observations for each of the beavers
points(beaver.1, col="darkred", stack=TRUE)
points(beaver.2, col="darkblue", stack=TRUE)
points(beaver.3, col="darkorange", stack=TRUE)

Visualization quick tips

Tip 1: The rose.diag function requires data in a circular format. This means the
data needs to be converted to circular class. When using your own data or other
datasets you can use the statement circular (your_data) to coerce data into a
circular format.

Ti p 2: You can change the number of bins by using the arguments bins in
rose.diag. The argument takes positive values.
Tip 3: Change the argument col in rose.diag to choose different bin colors. In the
points function set Stack =FALSE if you want individual observations plotted
around the edge of the outer circle.

Tip 4: To create a regular multivariable rose diagram do not enter the last three
lines of code.

Quick reference R code

install.packages("circular")
require(circular)
set.seed(1234)
beaver.1 <- rvonmises(n=50, mu=circular(0), kappa=3)
beaver.2 <- rvonmises(n=50, mu=circular(pi), kappa=3)
beaver.3 <- rvonmises(n=50, mu=circular(pi/2), kappa=3)
par(bg="cornsilk")
rose.diag(beaver.1, bins=18,axes=TRUE,ticks=TRUE,
border="black",col="darkred",tol=0,prop=1.5)
par(new=TRUE)
rose.diag(beaver.2, bins=18,axes=TRUE,ticks=TRUE,
border="black",col="darkblue",tol=0,prop=1.5)
par(new=TRUE)
rose.diag(beaver.3, bins=18,axes=TRUE,ticks=TRUE,
border="black",col="darkorange",tol=0,prop=1.5) points(beaver.1,
col="darkred", stack=TRUE)
points(beaver.2, col="darkblue", stack=TRUE)
points(beaver.3, col="darkorange", stack=TRUE)
NOTES

Write your notes below…

 36. Pollution rose

Figure 36 1 Carbon monoxide concentration measured at Marylebone

How to read the chart

The image shows the frequency of counts by wind direction for Carbon
monoxide concentration (parts per million) measured at Marylebone, London.
The petals are similar to the bars in a histogram and represent the frequency of
observations. The petals are color coded to capture the extent of concentration.
Circular bands are drawn at 5%,10%,15% and 20%. A petal that touches the
5% band represents 5% of the overall data.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")
require(openair)# this package contains the data and generates the plot

Step 2: Clean and prepare the data

data(mydata)#data frame contains data on Carbon monoxide concentration
Step 3: Plot the chart
pollutant <- c("co")# Carbon monoxide observations
# set title and subtitle
title=c("Carbon monoxide concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")

pollutionRose(mydata,paddle=FALSE,pollutant= pollutant,annotate=FALSE,
key.footer = pollutant, key.position = "right", key = TRUE, breaks =
8,main=title, sub=sub,auto.text=FALSE)

Visualization quick tips

Tip 1: The location of the key is determined by the argument key.posiition; It can
take values “left”,”right”,”top” and “bottom”. The number of colors (or breaks)
represented in the key is determined by the argument breaks; for many datasets
values between 6-8 work well.

Tip 2: You can change the plot to a modern wind rose by setting the argument
paddle=TRUE.

Tip 3: When substituting your own data note the pollutionRose function requires
data frames with a field "date" that can be in either POSIXct or Date format but
should be set to the Greenwich Mean Time time-zone. The data should be at
least hourly or higher resolution.

Tip 4: The chart used the argument pollutant <- c("co") to access the carbon
monoxide observations in the mydata data frame. This data frame contains a
number of other air quality observations taken in central London:
1. "ws" - Wind speed.
2. "wd" - Wind direction, in degrees from North.
3. "nox"- Oxides of nitrogen concentration.
4. "no2" - Nitrogen dioxide concentration.
5. "o3"- Ozone concentration.
6. "pm10" - Particulate PM10 fraction measurement.
7. "so2" - Sulfur dioxide concentration.
 8. "pm25" - Particulate PM2.5 fraction measurement.

Quick reference R code

install.packages("openair")
require(openair)
data(mydata)#dataframe contains the data on Carbon monoxide concentration, in
ppm, as a numeric vector
pollutant <- c("co")# Carbon monoxide data
title=c("Carbon monoxide concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")
pollutionRose(mydata,paddle=FALSE,pollutant= pollutant,annotate=F,
key.footer = pollutant, key.position = "right", key = TRUE, breaks =
8,main=title, sub=sub,auto.text=FALSE)
NOTES

Write your notes below…

 37. Polar frequency plot

Figure 37 1 Variation in Ozone concentration measured at Marylebone

How to read the chart

The image shows the variation by wind direction in Ozone concentration

measured at Marylebone, London. The rectangular colored bars are similar to
the bars in a histogram and represent the frequency of observations. The bars
are color coded to capture the extent of concentration. Circular bands are drawn
at intervals of 10%.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")
require(openair)# this package contains the data and generates the plot

Step 2: Clean and prepare the data

data(mydata)#dataframe contains data on Carbon monoxide concentration

Step 3: Plot the chart

pollutant <- c("o3")# Carbon monoxide observations
# set title and subtitle
title=c("Variation in Ozone concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")
polarFreq (mydata,paddle=FALSE,statistic="max",pollutant=
pollutant,annotate=F, key.footer = pollutant, key.position = "right", key =
TRUE,main=title,
sub=sub,auto.text=TRUE,grid.line=10,border.col="white",breaks = seq(0, 70,
10))

Visualization quick tips

Tip 1: The location of the key is determined by the argument key.posiition; It can
take values “left”,”right”,”top” and “bottom”.

Tip 2: The number of colors (or breaks) represented in the key is controlled by
the breaks argument in the polarFreq function. The imaged used breaks = seq(0,
70, 10), which generated a series of breaks from 0 to 70 in increments of 10.
Try breaks = seq(0, 70, 5) and observe the difference.

Tip 3: When substituting your own data note the polarFreq function requires
data frames with a field "date" that can be in either POSIXct or standard R Date
format. It should be set to the Greenwich Mean Time time-zone. The data should
ideally be at hourly or higher resolution.

Tip 4: The argument grid.line=10 controls the spacing of the dotted grid lines.
The argument border.col="white" controls the color of the boarder around each
colored rectangular. Set border.col = "transparent" to have an invisible border
around each colored rectangle.

Tip 5: The images used the argument pollutant <- c("o3") to access the carbon
monoxide observations in the mydata data frame. This data frame contains a
number of other air quality observations taken in central London:
 9. "ws" - Wind speed.
10. "wd" - Wind direction, in degrees from North.
11. "nox"- Oxides of nitrogen concentration.
12. "no2" - Nitrogen dioxide concentration.
13. "co"- Carbon monoxide concentration.
14. "pm10" - Particulate PM10 fraction measurement.
15. "so2" - Sulfur dioxide concentration.
16. "pm25" - Particulate PM2.5 fraction measurement.

Quick reference R code

install.packages("openair")
require(openair)
data(mydata)#dataframe contains the data on Carbon monoxide concentration, in
ppm, as a numeric vector
pollutant <- c("o3")# Carbon monoxide data
title=c("Variation in Ozone concentration in central London")
sub=c("1st January 1998 to 23rd June 2005")
polarFreq (mydata,paddle=FALSE,statistic="max",pollutant=
pollutant,annotate=F, key.footer = pollutant, key.position = "right", key =
TRUE,main=title,
sub=sub,auto.text=TRUE,grid.line=10,border.col="white",breaks = seq(0, 70,
10))
NOTES

Write your notes below…

 38. Australian wind rose

Figure 38 1 Wind direction & wind speed measured at Marylebone, London

How to read the chart

The chart shows a traditional wind rose using data measured at Marylebone,
London between 1st January 1998 to 23rd June 2005. Wind directions are
divided into eight compass directions. The circles around the image represent
the various percentages of occurrence of the winds. For example, the branch to
the north west just reaches the 10% ring implying a frequency of 10% blowing
from that direction. Calm has no direction.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")# contains the data
install.packages("plotrix")# used to draw the chart
require(openair)
require(plotrix)

Step 2: Clean and prepare the data

# get the wind speed and direction data
raw.data<-cbind(mydata$wd,mydata$ws)
# clean the data to remove missing observations
data<-na.omit(raw.data)
# put the contents into a table where the rows represent speed
#ranges and the columns indicating wind directions.
#also convert wind speed from m/s to km per hour
table<-bin.wind.records(data[,1],data[,2]*3.6,ndir=8)

Step 3: Plot the chart

par(bg="cornsilk")# set background color
oz.windrose(table)

Visualization quick tips

Tip 1: Data for the oz.windrose function needs to be in a format where the rows
represent speed ranges and the columns indicating wind directions. Furthermore
the data should be in percentages such that the sum of all data in your table is
equal to 100. This is best achieved using the bin.wind.records function. This
function classifies wind direction and speed records into a matrix of
percentages of observations in speed and direction bins. Use the argument ndir
to change the number of direction bins to be used in the wind rose. The
traditional number of bins is ndir=8.

Tip 2: Use the argument speed.col in the oz.windrose function to change the
color of the branches of the wind rose. The default colors, shown in the above
chart are obtained by speed.col
=c("#dab286","#fe9a66","#ce6733","#986434"). You could also set alternative
colors such as speed.col =c("red","green","purple","yellow").

Tip 3: To remove the legend add the argument show.legend=FALSE to the

oz.windrose function.
Quick reference R code

install.packages("openair")
install.packages("plotrix")
require(openair)
require(plotrix)

raw.data<-cbind(mydata$wd,mydata$ws)
data<-na.omit(raw.data)# remove missing observations
table<-bin.wind.records(data[,1],data[,2]*3.6,ndir=8)# create table and covert
wind speed to km per hour
par(bg="cornsilk")
oz.windrose(table)
NOTES

Write your notes below…

 39. Basic heat map

Figure 39 1 Violent crime in Pacific US States in 1973

How to read the chart

The chart is constructed from statistics, in arrests per 100,000 residents for
assault, murder, and rape in each of the five Pacific US states in 1973. The
shading goes from dark red to very light yellow. The darker the shading the
higher the number of violent offenses. The variable UrbanPop on the horizontal
axes refers to the size of the urban population, the darker the shading the higher
the urban population.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("datasets") # you only need to do this once
require(datasets) # datasets package contains the chart data.

Step 2: Clean and prepare the data

data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,],
USArrests[37,],USArrests[47,]))
# use data contained in data frame USArrests. Note the minus sign before the
rbind statement. This is set so when the chart is drawn in the next step, the
deeper the red (darker color) the higher the variable of interest. Alaska,
California, Hawaii, Oregon and Washington are in rows 2,5,11,37 and 47 of
USArrests.

Step 3: Plot the chart

heatmap(data, Rowv=NA,Colv = NA, scale="column",cexCol=1, cexRow=1
,col= heat.colors(10, alpha = 1) )

Visualization quick tips

Tip 1: The heatmap function requires data in a matrix format. When using your
own data or other datasets you can use the as.matrix function to coerce data into
a matrix.

Tip 2: You can change the size of the text on the column and row by using the
arguments cexRow, cexCol respectively. The arguments take positive values.
The larger the value, the larger the text. As a rule of thumb set both equal to 1
and then experiment for best effect.

Tip 3: The argument col=heat.colors takes two parameters. The first is the
number of colors to use. The second relates to the transparency of the colors.
The chart in this chapter used 10 colors. Try experimenting with a lower
number. The chart in this chapter set transparency equal to 1 (no transparency).
It takes values between 0 and 1. Try experimenting with different values, the
visual impact can be quite stunning.

Tip 4: The heat.colors plots the heatmap using shades of red, orange and
yellow. Many other options are available. For example terrain.colors(n, alpha =
1), topo.colors(n, alpha = 1) and cm.colors(n, alpha = 1). The parameter n
refers to the number of colors and alpha the degree of transparency. As an
illustration you might set col= terrain.colors(10, 1). Experiment with color and
parameter values to see if trends in the dataset appear more pronounced.

Quick reference R code

install.packages("datasets")
require(datasets)
data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,],
USArrests[37,],USArrests[47,]))
heatmap(data, Rowv=NA,Colv = NA, scale="column",cexCol=1, cexRow=1
,col= heat.colors(10, alpha = 1) )
NOTES

Write your notes below…

 40. Circular data heat map with dendrograms

Figure 40 1 Simulated direction of foraging ants at different times of the year

How to read the chart

The chart is constructed from simulated circular data representing the foraging
direction of eight ants from a single colony at various times of the year. The
shading in the main panel goes from dark red (higher circular values) to very
light yellow (lower circular values). The column dendrogram (top) shows the
relationship between the various months of the year. The row dendrogram (on
left of diagram) shows the relationship between the various ants.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular") # you only need to do this once
require(circular) # this package generates data and plot

Step 2: Clean and prepare the data

set.seed(1234) )# this allows you to replicate the chart
simulated.ant <- circular(matrix(rwrappedstable(48, scale=1, index =1,
skewness =0), 8,6) )
# a 8 by 6 matrix of simulated observations is stored in the variable
simulated.ant. The function rwrappedstable generates observations from a
wrapped stable distribution. An index value of 1 simulates from the Cauchy
distribution, an index value of 2 simulates from the normal distribution.
colnames(simulated.ant) <- c("Jan","Feb","Mar","Apr","May", "Jun") #create
column names
rownames(simulated.ant) <- c("ant 1","ant 2","ant 3","ant 4","ant 5","ant 6","ant
7","ant 8")#create row names

Step 3: Plot the chart

heatmap.circular (simulated.ant,cexCol=1,cexRow=1 ,col= heat.colors(10,
alpha = 1),ylab="single colony",xlab="six months of the year")

Visualization quick tips

Tip 1: The heatmap.circular function requires data in a matrix format. When

using your own data or other datasets you can use the as.matrix function to
coerce data into a matrix.

Tip 2: You can change the size of the text on the column and row by using the
arguments cexRow, cexCol respectively. The arguments take positive values.
The larger the value, the larger the text. As a rule of thumb set both equal to 1
and then experiment for best effect.

Tip 3: You can insert a title on to the chart by adding the argument main="my
title" to the heatmap.circular function. To suppress the row dendrogram add the
argument Rowv = NA to the heatmap.circular function. To suppress the column
dendrogram add the argument Colv = NA to the heatmap.circular function.

Tip 4: The argument col=heat.colors takes two parameters. The first is the
number of colors to use. The second relates to the transparency of the colors.
The chart in this chapter used 10 colors. Try experimenting with a lower number
of color. The chart in this chapter set transparency equal to 1 (no transparency).
It takes values between 0 and 1. Try experimenting with different values, the
visual impact can be quite stunning.

Tip 5: The heat.colors function plots shades of red, orange and yellow. Many
other options are available. For example terrain.colors(n, alpha = 1),
topo.colors(n, alpha = 1) and cm.colors(n, alpha = 1). The parameter n refers to
the number of colors and alpha the degree of transparency. As an illustration you
might set col= terrain.colors(10, 1). Experiment with color and parameter
values to see if trends in the dataset appear more pronounced.

Quick reference R code

install.packages("circular") # you only need to do this once

require(circular)
set.seed(1234)
simulated.ant <- circular(matrix(rwrappedstable(48, scale=1, index=1,
skewness =0), 8,6) )
colnames(simulated.ant) <- c("Jan","Feb","Mar","Apr","May","Jun")
rownames(simulated.ant) <- c("ant 1","ant 2","ant 3","ant 4","ant 5","ant 6","ant
7","ant 8")
heatmap.circular (simulated.ant,cexCol=1,cexRow=1 ,col= heat.colors(10,
alpha = 1),ylab="single colony",xlab="six months of the year")
NOTES

Write your notes below…

 41. Column trace dendrogram heat map

Figure 41 1 Violent crime in Pacific US States in 1973

How to read the chart

The chart is constructed from statistics, in arrests per 100,000 residents for
assault, murder, and rape in each of the five Pacific US states in 1973. The
shading goes from dark red to very light yellow. The darker the shading the
higher the number of violent offenses. The variable UrbanPop on the horizontal
axes refers to the size of the urban population, the darker the shading the higher
the urban population. The trace (solid line ) is drawn down the column's. The
distance of the line from the center of each color-cell is proportional to the size
of the measurement.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("datasets") # you only need to do this once
install.packages("gplots") # this package generates the plot
require(datasets) # datasets package contains the chart data.
require(gplots)
Step 2: Clean and prepare the data
data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,]
,USArrests[37,],USArrests[47,]))
# use data contained in data frame USArrests and set so the deeper the red
(darker color) the higher the variable of interest. Alaska, California, Hawaii,
Oregon and Washington are in rows 2,5,11,37 and 47 of USArrests.

Step 3: Plot the chart

heatmap.2(data,scale="column",cexCol=1,cexRow=1, col=heat.colors(10,
alpha = 1), density.info='none' , Rowv=NA,key=FALSE)

Visualization quick tips

Tip 1: The heatmap.2 function requires data in a matrix format. When using your
own data or other datasets you can use the as.matrix function to coerce data into
a matrix.

Tip 2: You can change the size of the text on the column and row by using the
arguments cexRow, cexCol respectively. The arguments take positive values.
The larger the value, the larger the text. As a rule of thumb set both equal to 1
and then experiment for best effect.

Tip 3: You can insert a title by adding the argument main="my title" to the
heatmap.2 function. To activate the row dendrogram add the argument Rowv =
TRUE. To suppress the column dendrogram add the argument Colv = NA.
Ti p 4: The argument col=heat.colors takes two parameters. The first is the
number of colors to use. The second relates to the transparency of the colors.
The chart in this chapter used 10 colors. Try experimenting with a lower number
of color. The chart in this chapter set transparency equal to 1 (no transparency).
It takes values between 0 and 1.
Tip 5: The heat.colors plots shades of red, orange and yellow. Many other
options are available. For example terrain.colors(n, alpha = 1), topo.colors(n,
alpha = 1) and cm.colors(n, alpha = 1). The parameter n refers to the number of
colors and alpha the degree of transparency. As an illustration you might set
col= terrain.colors(10, 1). Experiment with color and parameter values to see
if trends in the dataset appear more pronounced.

Tip 6: To add a color key to the chart set the argument key=TRUE. You can set
the size of the color key by using the keysize argument. It takes positive values.
The larger the value the larger the key. Try experimenting different values. Good
starting points are keysize=2 and keysize =1.

Quick reference R code

install.packages("datasets") # you only need to do this once

install.packages("gplots") # this package generates the plot
require(datasets) # datasets package contains the chart data.
require(gplots)
data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,],
USArrests[37,],USArrests[47,]))
heatmap.2(data, scale="column",cexCol=1,cexRow=1,col= heat.colors(10,
alpha = 1), density.info = 'none', Rowv=NA,key=FALSE)
NOTES

Write your notes below…

 42. Annotated heat map

Figure 42 1 Violent crime in Pacific US States in 1973

How to read the chart

The chart is constructed from statistics, in arrests per 100,000 residents for
assault, murder, and rape in each of the five Pacific US states in 1973. The
shading goes from dark red to very light yellow. The darker the shading the
higher the number of violent offenses. The variable UrbanPop on the horizontal
axes refers to the size of the urban population, the darker the shading the higher
the urban population. The chart has been annotated with comments on those
areas where the researcher has sufficient information to recommend policy
action (marked “yes”) and those areas where further information is sought
(market “no”).

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("datasets") # you only need to do this once
install.packages("gplots") # this package generates the plot
require(datasets) # datasets package contains the chart data.
require(gplots)

Step 2: Clean and prepare the data

data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,],
USArrests[37,],USArrests[47,]))
# use data contained in data frame USArrests and set so the deeper the red
(darker color) the higher the variable of interest. Alaska, California, Hawaii,
Oregon and Washington are in rows 2,5,11,37 and 47 of USArrests.
text<-matrix("",5,4) # Create a blank text matrix
text[1,1]= "yes" # add text comments to appear on heat plot
text[2,1]= "yes"
text[2,3]= "yes"
text[1,4]= "yes"
text[4,1]= "no"
text[3,2]= "no"
text[4,3]= "no"
text[3,4]= "no"

Step 3: Plot the chart

heatmap.2(data, scale="column",cexCol=1,cexRow=1 ,col= heat.colors(10,
alpha = 1), density.info='none', cellnote=text,
trace="none",notecol="darkblue",Rowv = FALSE,key=FALSE)

Visualization quick tips

Tip 1: The heatmap.2 function requires data in a matrix format. When using your
own data or other datasets you can use the as.matrix function to coerce data into
a matrix.

Tip 2: You can change the size of the text on the column and row by using the
arguments cexRow, cexCol respectively. The arguments take positive values.
The larger the value, the larger the text. As a rule of thumb set both equal to 1
and then experiment for best effect.
Tip 3: You can insert a title on to the chart by adding the argument main="my
title" to the heatmap.2 function. To activate the row dendrogram add the
argument Rowv = TRUE. To suppress the column dendrogram add the argument
Colv = NA.

Tip 4: The argument col=heat.colors takes two parameters. The first is the
number of colors to use. The second relates to the transparency of the colors.
The chart in this chapter used 10 colors. Try experimenting with a lower number
of color. The chart in this chapter set transparency equal to 1 (no transparency).
It takes values between 0 and 1. Try experimenting with different values, the
visual impact can be quite stunning.

Tip 5: The heat.colors plots shades of red, orange and yellow. Many other
options are available. For example terrain.colors(n, alpha = 1), topo.colors(n,
alpha = 1) and cm.colors(n, alpha = 1). The parameter n refers to the number of
colors and alpha the degree of transparency. As an illustration you might set
col= terrain.colors(10, 1). Experiment with color and parameter values to see
if trends in the dataset appear more pronounced.

Tip 6: To add a color key to the chart set the argument key=TRUE. You can set
the size of the color key by using the keysize argument. It takes positive values,
larger values increase the size of the key. Try experimenting different values.
Good starting points are keysize=2 and keysize =1.

Quick reference R code

install.packages("datasets")
install.packages("gplots")
require(datasets)
require(gplots)
data<-as.matrix(-rbind(USArrests[2,],USArrests[5,],USArrests[11,],
USArrests[37,],USArrests[47,]))
text<-matrix("",5,4)
text[1,1]= "yes"
text[2,1]= "yes"
text[2,3]= "yes"
text[1,4]= "yes"
text[4,1]= "no"
text[3,2]= "no"
text[4,3]= "no"
text[3,4]= "no"
heatmap.2(data, scale="column", cexCol=1,cexRow=1 ,col= heat.colors(10,
alpha = 1), density.info='none', cellnote=text,
trace="none",notecol="darkblue",Rowv = FALSE,key=FALSE)
NOTES

Write your notes below…

 43. Correlation matrix

Figure 43 1 Air quality factor correlations measured at Marylebone

How to read the chart

The chart is constructed from the correlation between the eight air quality
variables collected at the air quality monitoring station in Marylebone, London.
The actual correlations are printed inside each square. The grayscale shading of
each square represents the strength of correlation. The scale runs from black
(negative 100% correlation) to white (positive 100% correlation).

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot

Step 2: Clean and prepare the data

# calculate the correlation between indicators
x<-na.omit(mydata[,2:9])#remove missing values
cor<- cor(x)*100 # make sure correlation is between -100 to +100

Step 3: Plot the chart

par(bg="cornsilk")# set background color
color2D.matplot(x=round(cor,1), show.values=TRUE,axes=FALSE,
xlab="",ylab="")
#add axis
axis(side=1,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
axis(side=2,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
#add titles
title("London Air Quality ",col.main="darkblue",line=2, cex.main =2)
title("January 1998- June 2005",col.main="black",line=1, cex.main =1)
title("*correlations calculated from hourly
measurements",col.main="black",line=-29, cex.main =0.8)

Visualization quick tips

Tip 1: You can replace the squares with hexagons by adding the argument
do.hex=TRUE to the plotting function color2D.matplot.

Tip 2: To hide the correlation values set the argument show.values=FALSE.

You can also supply your own x and y axis labels by setting xlab="my x axis
label",ylab="my y axis label".

Tip 3: To add a graded legend to the chart add the argument

show.legend=TRUE to the plotting function color2D.matplot.

Quick reference R code

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot
x<-na.omit(mydata[,2:9])#remove missing values
cor<- cor(x)*100
par(bg="cornsilk")
color2D.matplot(x=round(cor,1), show.values=TRUE,axes=FALSE,
xlab="",ylab="")
#create basic plot area
axis(side=1,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
axis(side=2,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
title("London Air Quality ",col.main="darkblue",line=2, cex.main =2)
title("January 1998- June 2005",col.main="black",line=1, cex.main =1)
title("*correlations calculated from hourly
measurements",col.main="black",line=-29, cex.main =0.8)
NOTES

Write your notes below…

 44. Hinton plot

Figure 44 1 Derived Hinton plot from London air quality data

How to read the chart

The chart is constructed from the correlation between eight air quality variables
collected at the air quality monitoring station in Marylebone, London. The size
of the squares are proportional to the correlation with the sign indicated by the
color of the squares.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot

Step 2: Clean and prepare the data

# calculate the correlation between indicators
x<-na.omit(mydata[,2:9])#remove missing values
cor<- cor(x)*100 # make sure correlation is between -100 to +100

Step 3: Plot the chart

par(bg=" tan")# set background color
color2D.matplot(x=round(cor,1), Hinton=T, show.values=F,axes=FALSE,
xlab="",ylab="",show.legend=T,extremes=c(2,3))
#create basic plot area
axis(side=1,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
axis(side=2,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
#add titles
title("Hinton Air Quality Map",col.main="purple",line=2, cex.main =2)
title("Central London",col.main="purple",line=1, cex.main =1)
title("*Hourly measurements taken between January 1998- June
2005",col.main="black",line=-27.9, cex.main =0.6)

Visualization quick tips

Tip 1: You can create a correlation matrix by setting Hinton =FALSE in the
plotting function color2D.matplot. Replace the squares with hexagons by adding
the argument do.hex=TRUE to the plotting function color2D.matplot.

Tip 2: To see the actual variable values set the argument show.values=TRUE.
You can also supply your own x and y axis labels by setting xlab="my x axis
label",ylab="my y axis label".

Tip 3: To remove the legend set the argument show.legend=FALSE in the

plotting function color2D.matplot.

Quick reference R code

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot
x<-na.omit(mydata[,2:9])#remove missing values
cor<- cor(x)*100
par(bg=" tan")
color2D.matplot(x=round(cor,1), Hinton=T, show.values=F,axes=FALSE,
xlab="",ylab="",show.legend=T,extremes=c(2,3))
axis(side=1,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
axis(side=2,at=(0.7:7.7), labels=colnames(cor),cex.axis=1, tick=FALSE)
title("Hinton Air Quality Map",col.main="purple",line=2, cex.main =2)
title("Central London",col.main="purple",line=1, cex.main =1)
title("*Hourly measurements taken between January 1998- June
2005",col.main="black",line=-27.9, cex.main =0.6)
NOTES

Write your notes below…

 45. Battleship plot

Figure 45 1 Battleship plot of Swiss data collected during 1888

How to read the chart

The chart is constructed from the correlation between the six socioeconomic
indicators collected in French-speaking provinces of Switzerland around 1888.
The size of the rectangular bars indicate the strength of the correlation. The
larger the bar the higher the absolute correlation between two indicators. The
bars are color coded to help distinguish positive correlated indicators from
negatively correlated indicators.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("plotrix")#package will draw plot
require(plotrix)
require(datasets)# contains the Swiss data
Step 2: Clean and prepare the data
# calculate the correlation between indicators
data<-cor(swiss)
#set the diagonal elements of the correlation matrix to zero
data[1,1]=0
data[2,2]=0
data[3,3]=0
data[4,4]=0
data[5,5]=0
data[6,6]=0

Step 3: Plot the chart

par(bg="burlywood1")# set background color
battleship.plot(data,col=color.scale(-data,c(0,300),70,60, color.spec="hcl"),
cex.labels=0.8,border="burlywood1")
#add title
title("Swiss Socioeconomic Indicators",col.main="brown",line=3.5, cex.main
=1.4)

#add legend
legend(x=5.5,y=1.3, inset=.05, bty="n", cex=0.75, text.col ="darkred",c("Strong
Pos", "Moderate Pos", "Moderate Neg","Strong Neg"), fill= c("#D97088FF",
"#A49100FF", "#00AA83FF", "#B678D5FF"))

Visualization quick tips

Tip 1: The argument border="burlywood1" ensures the diagonal elements of the

diagram are invisible. If you wish to see them, and have a border around each
rectangle set border=“your color”.

Tip 2: You can change the size of the text on the x and y axis by using the
argument cex.labels. The default value is 1, larger values increase the size of the
x,y axis labels.

Tip 3: The function title is a simple, fast and flexible way to add a title to a
chart. Use the line argument to place the title precisely where you require it.

Quick reference R code

install.packages("plotrix")#package will draw plot
require(plotrix)
require(datasets)

data<-cor(swiss)
#set the diagonal elements to zero
data[1,1]=0
data[2,2]=0
data[3,3]=0
data[4,4]=0
data[5,5]=0
data[6,6]=0
par(bg="burlywood1")
battleship.plot(data,col=color.scale(-data,c(0,300),70,60, color.spec="hcl"),
cex.labels=0.8,border="burlywood1")

title("Swiss Socioeconomic Indicators",col.main="brown",line=3.5, cex.main

=1.4)
legend(x=5.5,y=1.3, inset=.05, bty="n", cex=0.75, text.col ="darkred",c("Strong
Pos", "Moderate Pos", "Moderate Neg","Strong Neg"), fill= c("#D97088FF",
"#A49100FF", "#00AA83FF", "#B678D5FF"))
NOTES

Write your notes below…

 46. Density rose split plot

Figure 46 1 Three groups of desert ants –density plot with rose diagram

How to read the chart

The panel displays the circular density and rose diagram for the orientation of
three sets of long-legged desert ants after one eye on each ant was 'trained' to
learn the ant's home direction, then covered and the other eye uncovered. The
top three images show the density fit for each group to the circular distribution.
The bottom three images present the rose diagrams for each group of ants. The
size of the “petals” indicate the primary orientation of each group of ants.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
# The data frame fisherB10 contains observations on each of the ant groups. It is
converted to an object of class circular and orientated so that zero equals pi/2.
This will ensure zero appears at the top of the resultant arrow plots.

Step 3: Plot the chart

# create a 2 by 3 area, each of the six charts is plotted on one area and set the
background color to cornsilk.
par(mfrow=c(2,3), bg="cornsilk")par(bg="cornsilk")
# fit density and then create plot for first group of ants
density.1<- density(ant.1, bw=5)
plot(density.1, points.plot=TRUE, col="darkred",main="Ant group 1",lwd=2)
# fit density and then create plot for second group of ants
density.2<- density(ant.2, bw=5)
plot(density.2, points.plot=TRUE, col="darkgreen",main="Ant group 2")
# fit density and then create plot for third group of ants
density.3<- density(ant.3, bw=5)
plot(density.3, points.plot=TRUE, col="darkblue",main="Ant group 3")
# create and plot the rose diagram for first group of ants
rose.diag(ant.1, bins=15,axes=TRUE, ticks=TRUE ,border=
"black",col="darkred",tol=0,prop=1.5, main="Ant group 1")
# create and plot the rose diagram for second group of ants
rose.diag(ant.2, bins=15,axes=TRUE, ticks=TRUE ,border=
"black",col="darkgreen",tol=0,prop=1.5, main="Ant group 2")
# create and plot the rose diagram for third group of ants
rose.diag(ant.3, bins=15,axes=TRUE, ticks=TRUE ,border=
"black",col="darkblue",tol=0,prop=1.5, main="Ant group 3")

Visualization quick tips

Tip 1: The rose.diag function require data in a circular format. This means the
data needs to be converted to circular class. When using your own data or other
datasets you can use the statement circular (your_data) to coerce data into a
circular format.
Ti p 2: You can exclude the axes on each chart by setting the argument
axes=FALSE in the plot function.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Tip 4: Change the plot region using par(mfrow=c(x,y)), where x is the number
of rows, and y is the number of columns. For example, par(mfrow=c(2,2))
would split the graphic output into four areas and then you could plot a chart to
each specific area.

Quick reference R code

install.packages("circular") # this package contains the data and generates the

plot
require(circular)
ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
par(mfrow=c(2,3), bg="cornsilk")
density.1<- density(ant.1, bw=5)
plot(density.1, points.plot=TRUE, col="darkred",main="Ant group 1",lwd=2)
density.2<- density(ant.2, bw=5)
plot(density.2, points.plot=TRUE, col="darkgreen",main="Ant group 2")
density.3<- density(ant.3, bw=5)
plot(density.3, points.plot=TRUE, col="darkblue",main="Ant group 3")

rose.diag(ant.1, bins=15,axes=TRUE,ticks=TRUE
,border="black",col="darkred",tol=0,prop=1.5, main="Ant group 1")
rose.diag(ant.2, bins=15,axes=TRUE,ticks=TRUE
,border="black",col="darkgreen",tol=0,prop=1.5, main="Ant group 2")
rose.diag(ant.3, bins=15,axes=TRUE,ticks=TRUE
,border="black",col="darkblue",tol=0,prop=1.5, main="Ant group 3")
NOTES

Write your notes below…

 47. Cylindrical graded bar plot

Figure 47 1 Rural and urban death rates in Virginia (1940)

How to read the chart

The chart displays the death rate per 1000 in Virginia during 1940. The data are
broken out by gender and by location (urban, rural) The cylindrical bars are
shaded by age group.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("plotrix")#used to draw the plot
require(plotrix)
require(datasets)#contains the data used in the example

Step 2: Clean and prepare the data

age<-c(54,59,64,69,74)#used to color bars
names=colnames(VADeaths) #used to label x-axis

Step 3: Plot the chart

main= "Death rates per 1000 in Virginia in 1940"# title
par(bg="lightskyblue1") #background color
barp(VADeaths,col=color.scale(age,c(1.1,0),c(0,1.1),c(0,0.1)),cylindrical=TRUE,names
# text for legend
age.range=c("50-54", "55-59","60-64", "65-69", "70-74")
#create legend
legend(x="topright", inset=.05, bty="n", cex=0.75, text.col ="darkred",
legend=age.range, fill= c("#FF0000FF", "#BF4006FF", "#80800DFF",
"#40BF13FF", "#00FF1AFF"), title="Age Range")

Visualization quick tips

Tip 1: You can draw standard rectangular bars by setting the argument
cylindrical=FALSE.

Tip 2: It is possible to add colors to the legend plot by specifying the color
name rather than number, for example fill= c("red", "orange", "purple", "green",
"blue")

Tip 3: Change the size of the text in the legend by using the argument cex. It has
a default value of 1, larger values increase the text size.

Quick reference R code

install.packages("plotrix")
require(plotrix)
require(datasets)
age<-c(54,59,64,69,74)
names=colnames(VADeaths)
main= "Death rates per 1000 in Virginia in 1940"
par(bg="lightskyblue1")
barp(VADeaths,col=color.scale(age,c(1.1,0),c(0,1.1),c(0,0.1)),cylindrical=TRUE,names
age.range=c("50-54", "55-59","60-64", "65-69", "70-74")
legend(x="topright", inset=.05, bty="n", cex=0.75, text.col ="darkred",
legend=age.range, fill= c("#FF0000FF", "#BF4006FF", "#80800DFF",
"#40BF13FF", "#00FF1AFF"), title="Age Range")
NOTES

Write your notes below…

 48. Centipede plot

Figure 48 1 Variation in air quality metrics measured at Marylebone

How to read the chart

The chart displays 48 hour variation in various air quality measurements taken
in central London from January 1st 1998 to June 23rd 2005. The extent of the red
line represents captures the degree of variation of each environmental factor;
The blue square in the middle of each line represents the median or average
value.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot

Step 2: Clean and prepare the data

#mydata contains the observations used
x<-na.omit(mydata[,2:9])#remove missing values
diff=48# used to calculate 48 hour change
#calculate the 48 hour difference of each variable
x1 <- diff(x[,1], k= diff)
x2 <- diff(x[,2], k= diff)
x3 <- diff(x[,3], k= diff)
x4 <- diff(x[,4], k= diff)
x5 <- diff(x[,6], k= diff)
x6 <- diff(x[,7], k= diff)
x7 <- diff(x[,8], k= diff)

#divide 48 hour changes by the maximum 48 hour change

x1 <- x1/max(x1)
x2 <- x2/max(x2)
x3 <- x3/max(x3)
x4 <- x4/max(x4)
x5 <- x5/max(x5)
x6 <- x6/max(x6)
x7 <- x7/max(x7)
#create a list suitable for use in the plot function
temp<-list("",7)
temp[[1]]<-x1
temp[[2]]<-x2
temp[[3]]<-x3
temp[[4]]<-x4
temp[[5]]<-x5
temp[[6]]<-x6
temp[[7]]<-x7
# put data into correct format for plotting
data<-get.segs(temp)
# add column names
colnames(data)=c("ws", "wd", "nox", "no2" , "pm10", "so2", "co")
Step 3: Plot the chart
par(bg="cornsilk")# set background color
centipede.plot(data*10000,sort.segs=TRUE,right.labels=rep("",7),main="Air
Quality Data - London",xlab="(48 hour change/ maximum 48 hour
change)",pch=22,col="red",bg="blue",col.main="blue")

Visualization quick tips

Tip 1: Change the background color of the plot using par(bg="my color") .

Tip 2: The observations are sorted by average value, this is controlled by the
argument sort.segs=TRUE. Set to FALSE to plot in the order they appear in the
dataset.

Tip 3: Remove the argument right.labels=rep("",7) if you would like to see a

count of the number of observations for each variable appear on the right hand
side of the graph.

Tip 4: The arguments pch,col,bg, and col.main control the type of shape used to
represent the median (use values in the range 19-26), the color of the lines
representing variation, color of the shape used to represent the median and the
color used in the title.

Quick reference R code

install.packages("openair")
install.packages("plotrix")
require(openair)#contains the data
require(plotrix)# draws the plot

x<-na.omit(mydata[,2:9])#remove missing values

diff=48# used to calculate 48 hour change

x1 <- diff(x[,1], k= diff)
x2 <- diff(x[,2], k= diff)
x3 <- diff(x[,3], k= diff)
x4 <- diff(x[,4], k= diff)
x5 <- diff(x[,6], k= diff)
x6 <- diff(x[,7], k= diff)
x7 <- diff(x[,8], k= diff)

#divide 48 hour changes by the maximum 48 hour change

x1 <- x1/max(x1)
x2 <- x2/max(x2)
x3 <- x3/max(x3)
x4 <- x4/max(x4)
x5 <- x5/max(x5)
x6 <- x6/max(x6)
x7 <- x7/max(x7)

temp<-list("",7)
temp[[1]]<-x1
temp[[2]]<-x2
temp[[3]]<-x3
temp[[4]]<-x4
temp[[5]]<-x5
temp[[6]]<-x6
temp[[7]]<-x7
data<-get.segs(temp)
colnames(data)=c("ws", "wd", "nox", "no2" , "pm10", "so2", "co")
par(bg="cornsilk")
centipede.plot(data*10000,sort.segs=TRUE,right.labels=rep("",7),main="Air
Quality Data - London",xlab="(48 hour change/ maximum 48 hour
change)",pch=22,col="red",bg="blue",col.main="blue")
NOTES

Write your notes below…

 49. Circular probability split plot

Figure 49 1 Probability and empirical plots for three groups of desert ants

How to read the chart

The chart displays probability and empirical probability plots for the
orientation of three sets of long-legged desert ants after one eye on each ant was
'trained' to learn the ant's home direction, then covered and the other eye
uncovered. The top three images plot the probability plot for each group relative
to the von Misses distribution. The closer the dotted points are the solid black
line, the better the observations fit the von Misses distribution. The bottom three
charts plot the empirical cumulative probability distribution of the three groups
of ants.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("circular")
require(circular)# this package contains the data and generates the plot
require(circular)

Step 2: Clean and prepare the data

ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
# The data frame fisherB10 contains observations on each of the ant groups. It is
converted to an object of class circular and orientated so that zero equals pi/2.
This will ensure zero appears at the top of the resultant arrow plots.

Step 3: Plot the chart

# create a 2 by 3 area, each of the six charts is plotted on one area
par(mfrow=c(2,3))
par(bg="cornsilk")# set the background color to cornsilk
# probability plots of each group of ants
pp.plot(ant.1,col="darkred",pch=19,main="Ant group 1",)
pp.plot(ant.2,col="darkgreen",pch=19,main="Ant group 2")
pp.plot(ant.3,col="darkblue",pch=19,main="Ant group 3")
#cumulative probability plots of each group of ants
plot.edf(ant.1,col="darkred",pch=19,main="Ant group 1", ylab="Probability",
xlab="value")
plot.edf (ant.2,col="darkgreen",pch=19,main="Ant group 2",
ylab="Probability", xlab="value")
plot.edf (ant.3,col="darkblue",pch=19,main="Ant group 3", ylab="Probability",
xlab="value")

Visualization quick tips

Tip 1: The pp.plot and plot.edf function require data in a circular format. This
means the data needs to be converted to circular class. When using your own
data or other datasets you can use the statement circular (your_data) to coerce
data into a circular format.

Tip 2: You can include axes on each chart by setting the argument axes=TRUE
in the plot function.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Tip 4: Change the plot region using par(mfrow=c(x,y)), where x is the number
of rows, and y is the number of columns. For example, par(mfrow=c(2,2))
would split the graphic output into four areas and then you could plot a chart to
each specific area.

Quick reference R code

install.packages("circular")
require(circular)
ant.1<-circular(fisherB10$set1,zero=pi/2)
ant.2<-circular(fisherB10$set2,zero=pi/2)
ant.3<-circular(fisherB10$set3,zero=pi/2)
par(mfrow=c(2,3))
par(bg="cornsilk")
pp.plot(ant.1,col="darkred",pch=19,main="Ant group 1",)
pp.plot(ant.2,col="darkgreen",pch=19,main="Ant group 2")
pp.plot(ant.3,col="darkblue",pch=19,main="Ant group 3")
plot.edf(ant.1,col="darkred",pch=19,main="Ant group 1", ylab="Probability",
xlab="value")
plot.edf (ant.2,col="darkgreen",pch=19,main="Ant group 2",
ylab="Probability", xlab="value")
plot.edf (ant.3,col="darkblue",pch=19,main="Ant group 3", ylab="Probability",
xlab="value")
NOTES

Write your notes below…

 50. Wind - pollution stacked bar time series
chart

Figure 50 1 Annual ozone concentration by wind direction taken in London

How to read the chart

The chart displays the annual (median) distribution of ozone concentration in

central London by direction of the prevailing wind over the years 1998 to 2005.
The key on the right side of the graph gives the color codes for wind direction
(wd on chart), with dark blue representing “North”.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")# this package will draw the plot
require(openair)

Step 2: Clean and prepare the data

data(mydata)#dataframe contains the data on ozone concentrations
Step 3: Plot the chart
# text for the title
title="Ozone concentration by wind direction in London"
timeProp(mydata, pollutant="o3", avg.time="year", proportion="wd",normalise
= TRUE ,cols="jet", box.width=.9, key.position="right", statistic =
"frequency",xlab= "Year",main=title,cex.main=1.3)

Visualization quick tips

Tip 1: Change the color of the stacked bars by using the argument cols= "my
color". Quick and easy alternatives include “default”, “increment” and “heat”.

Tip 2: Set the argument normalise = FALSE to plot the actual frequencies in
place of the proportions.

Tip 3: The argument box.width determines the size of the gap between colored
bars. It takes a default value of 1 (no gap). Values less than one increase the
gap/space between the colored bars.

Tip 4: The argument key.position determines the location of the legend. It can
take values "right", "left", "top", "bottom".

Quick reference R code

install.packages("openair")
require(openair)
data(mydata)
title="Ozone concentration by wind direction in London"
timeProp(mydata, pollutant="o3", avg.time="year", proportion="wd",normalise
= TRUE ,cols="jet",box.width=.9,key.position="right", statistic =
"frequency",xlab= "Year",main=title,cex.main=1.3)
NOTES

Write your notes below…

 51. Environmental style time series summary
plot

Figure 51 1 Summary plot of air quality data

How to read the chart

The top panel displays ozone concentration (grey line) in central London over
the years 1998 to 2005. The cyan bar directly underneath the ozone time series
indicates the presence (cyan) or absence (black) of data at a particular point in
time. The box to the right of the top panel presents the density curve (red) of
ozone. The remaining four panels (nitrogen dioxide concentration, Oxides of
nitrogen concentration, wind direction and wind speed) have a similar
interpretation.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("openair")# this package will draw the plot
require(openair)

Step 2: Clean and prepare the data

data(mydata)#dataframe contains the data on ozone concentrations
temp<-mydata
data <- temp[,c(1:6)] # select the data used in chart
Step 3: Plot the chart
# text for the title
title="London Air Quality Data"
summaryPlot(data,type="density",col.hist="red",col.trend="grey",col.data="lightblue",co

Visualization quick tips

Tip 1: Change the color of the line used to draw the density, trend pollution
values and present and missing data by setting to “my color” the arguments
col.hist,col.trend,col.data and col.mis respectively.

Tip 2: For greater clarity plot fewer time series. This can be achieved by using
data <- temp[,c(1:number)], where number is equal to the number of time series
you wish to observe. You can plot all of the data in myframe by setting data<-
myframe in the above R code.

Tip 3: To use default labels for the x-axis omit the argument xlab. To add
custom text to the y-axis use the argument ylab="my text".

Quick reference R code

install.packages("openair")
require(openair)
data(mydata)#dataframe contains the data on ozone concentrations
temp<-mydata
data <- temp[,c(1:6)]
title="London Air Quality Data"
summaryPlot(data,type="density",col.hist="red",col.trend="grey",col.data="lightblue",co
NOTES

Write your notes below…

 52. Bumps chart

Figure 52 1 Rank position of original county cricket teams over time

How to read the chart

The chart displays the eight counties that took part in the first English cricket
country championships beginning in 1890. Surrey won the first championship
and are the first team listed under 1890. Sussex scored the lowest number of
points in that first season and are therefore placed last for 1890. The lines from
each team represent their relative ranking in both 1948 and 2012 using total
points earned.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("plotrix")# package will draw plot
require(plotrix)

Step 2: Clean and prepare the data

#create a matrix of league scores for 1890,1948 and 2012
points<-matrix(c(6,168,139,4,152,106,3,76,170,31,56,194,0,80,163,-
1,128,126,-5,160,172,-10,72,167),ncol=3,byrow=TRUE)
#add row names
rownames (points) <-c("Surrey","Lancashire","Kent","Yorkshire",
"Nottinghamshire","Gloucestershire","Middlesex","Sussex")
#add dates to columns
colnames(points)<-c("1890","1948","2012")
Step 3: Plot the chart
par(bg="white")# set the background color
bumpchart(points,rank=TRUE,col=rainbow(8),main="County Championship's
Original Eight Ranking Over Time")

Visualization quick tips

Tip 1: The color of the lines was set using col=rainbow(8), you can also specify
the colors of your choice directly by using col=c("your first color","your
second color"…,"your final color"). For example, try
col=c("red","green","yellow","purple","blue","orange","darkred","white")

Tip 2: The argument rank=TRUE ranks the values before plotting. To turn this
option off use rank =FALSE.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Tip 4: The first two arguments in the legend function, determine the horizontal
and vertical location of the legend. An alternative, which works in many
instances is to specify the location such as legend("topleft",…). You can also
specify "topright", "bottomright", "bottomleft".

Quick reference R code

install.packages("plotrix")# package will draw plot

require(plotrix)
points<-matrix(c(6,168,139,4,152,106,3,76,170,31,56,194,0,80,163,-
1,128,126,-5,160,172,-10,72,167),ncol=3,byrow=TRUE)
rownames (points) <-c("Surrey","Lancashire","Kent","Yorkshire",
"Nottinghamshire","Gloucestershire","Middlesex","Sussex")
colnames(points)<-c("1890","1948","2012")
par(bg="white")
bumpchart(points,rank=TRUE,col=rainbow(8),main="County Championship's
Original Eight Ranking Over Time")
NOTES

Write your notes below…

 53. Two ordinate time series stack plot

Figure 53 1 Stock – bond allocation with asset class returns

How to read the chart

The chart displays the allocation between bonds (blue) and stocks (tan) for a
portfolio over the years 2006 to 2013. The left axis provides the measurements
for the allocation. Overlaid on the plot are the annual returns for private equity,
real estate and gold with return values measured on the right axis.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("plotrix")# package will draw plot
require(plotrix)

Step 2: Clean and prepare the data

#enter stock and bond allocation
stock.percent =c(90,95,80,70,60,65,70,70)
bond.percent =100-stock.percent
# enter asset class returns
private_equity=c(31.2,21.4,-25.4,10.5,17,12.8,8,11.9)
real_estate=c(5.6,18.1,2.3,-35.7,6.1,7,5,9.3)
gold=c(23.2,31.3,7.5,25.3,28.7,13,0,-10.7)
#create a variable to count time for the basic plot
time<-0:7
# combine stocks and bond allocation
allocation <-cbind(bond.percent,stock.percent)
#combine asset class returns
return<-cbind(private_equity, real_estate, gold)

Step 3: Plot the chart

#set background and axis colors
par(bg=" lightsteelblue3",col=" lightsteelblue3", col.axis=" lightsteelblue3")
#create the basic plot
twoord.stackplot(lx=time, rx=time, ldata=allocation,
rdata=return, lcol=c("darkblue", "tan"),
rcol=c("brown", "red","cyan"), ltype="bar", rtype=c("o","o","o"),
lylab="Allocation (%)", rylab="Asset returns (%)", xlab="Year",
incrylim=-2/100,main="Portfolio Allocation and Returns")
par(bg=" lightsteelblue3",col="black", col.axis="black")
#create fancy axis
axis(side=1,at=(0:7), labels=
c("2006","2007","2008","2009","2010","2011","2012","2013"))

par(xpd=TRUE) #extend the area of plotting

par(new=TRUE) #to add new graph "layers"
plot(0:1, 0:1, type="n", xlab="",ylab="", axes=FALSE) #redo the (x,y) range

#asset allocation legend

legend(-0.18, 1.2, inset=.05,
bty="n", cex=1, text.col ="darkred",
c("Bonds", "Stocks"), fill=c("darkblue", "tan"))

#asset classes legend

legend(0.8, 1.2, inset=.05,
bty="n", cex=1, text.col ="darkred",
c("Private Equity", "Real Estate", "Gold"), fill= c("brown", "red","cyan"))

par(xpd=FALSE, new=FALSE) #default setting

Visualization quick tips

Tip 1: Both the ltype and rtype arguments can take various values. The most
frequently used are "p" for points, "l" for lines,"b" for both, "c" for the lines part
alone of "b", "o" for both ‘overplotted’, "h","s" for stair steps, "S" for other
steps, and "n" for no plotting.

Tip 2: You can change the right hand axis color specifying it as the first value in
the rcol argument i.e. rcol =c("my color","next color"…). Similarly, you can
change the left hand axis color using the same approach.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Tip 4: The first two argument in the legend function, determine the horizontal
and vertical location of the legend. An alternative, which works in many
instances is to specify the location such as legend("topleft",…). You can also
specify "topright", "bottomright", "bottomleft".

Quick reference R code

install.packages("plotrix")
require(plotrix)

stock.percent =c(90,95,80,70,60,65,70,70)
bond.percent =100-stock.percent
private_equity=c(31.2,21.4,-25.4,10.5,17,12.8,8,11.9)
real_estate=c(5.6,18.1,2.3,-35.7,6.1,7,5,9.3)
gold=c(23.2,31.3,7.5,25.3,28.7,13,0,-10.7)

time<-0:7

allocation <-cbind(bond.percent,stock.percent)

return<-cbind(private_equity, real_estate, gold)

par(bg=" lightsteelblue3",col=" lightsteelblue3", col.axis=" lightsteelblue3")

twoord.stackplot(lx=time, rx=time, ldata=allocation,
rdata=return, lcol=c("darkblue", "tan"),
rcol=c("brown", "red","cyan"), ltype="bar", rtype=c("o","o","o"),
lylab="Allocation (%)", rylab="Asset returns (%)", xlab="Year",
incrylim=-2/100,main="Portfolio Allocation and Returns")
par(bg=" lightsteelblue3",col="black", col.axis="black")

axis(side=1,at=(0:7), labels=
c("2006","2007","2008","2009","2010","2011","2012","2013"))

par(xpd=TRUE) #extend the area of plotting

par(new=TRUE) #to add new graph "layers"
plot(0:1, 0:1, type="n", xlab="",ylab="", axes=FALSE) #redo the (x,y) range
legend(-0.18, 1.2, inset=.05,
bty="n", cex=1, text.col ="darkred",
c("Bonds", "Stocks"), fill=c("darkblue", "tan"))
legend(0.8, 1.2, inset=.05,
bty="n", cex=1, text.col ="darkred",
c("Private Equity", "Real Estate", "Gold"), fill= c("brown", "red","cyan"))
par(xpd=FALSE, new=FALSE) #default setting
NOTES

Write your notes below…

 54. Two ordinate time series plot

Figure 54 1 Wind and temperature measurements taken in New York city

How to read the chart

The chart displays wind (solid green) and air temperature (purple line) taken in
New York during 1973. The left axis provides the measurements for wind, and
the right axis the measurements for temperature.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("plotrix")# package will draw plot
require(plotrix)
attach(airquality)#contains the dataset

Step 2: Clean and prepare the data

x=airquality[,3]#Wind measurement
y= airquality[,4]#temperature
seq<-1:153#used for x axis
labels<-c("May ","June","Aug ","Sept")# dates plotted on image
Step 3: Plot the chart
par(bg="burlywood1")#set background color
twoord.plot(seq,x,y,type=c("h","l"),rcol="purple",lcol="darkgreen",lwd=2,ylab=
"Wind measurement",rylab=" Temperature",main= "Atmospheric
Measurements",xticklab=labels,xtickpos =c(0,50,100,150),sub=" New York
1973")

Visualization quick tips

Tip 1: Change type argument to draw different lines. It takes a range of values.
The most frequently used are "p" for points, "l" for lines,"b" for both, "c" for the
lines part alone of "b", "o" for both ‘overplotted’, "h","s" for stair steps, "S" for
other steps, and "n" for no plotting.

Tip 2: You can change the right hand axis color by using the argument rcol="my
color". Similarly, you can change the left hand axis color by lcol="my color"

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("plotrix")
require(plotrix)
attach(airquality)
x=airquality[,3]#Wind measurement
y= airquality[,4]#temperature
seq<-1:153
labels<-c("May ","June","Aug ","Sept")
par(bg="burlywood1")
twoord.plot(seq,x,y,type=c("h","l"),rcol="purple",lcol="darkgreen",lwd=2,ylab=
"Wind measurement",rylab=" Temperature",main= "Atmospheric
Measurements",xticklab=labels,xtickpos=c(0,50,100,150),sub=" New York
1973")
NOTES

Write your notes below…

 55. Chernoff faces

Figure 55 1 Ice cream consumption in the United States

How to read the chart

The image displays Chernoff faces calculated from four variables, consumption
of ice cream per head (in pints); average family income per week (in US
Dollars); price of ice cream (per pint) and the average temperature (in
Fahrenheit). The variables used to define the face were height of face, width of
face, structure of face, height of mouth, width of mouth , degree of smiling,
height of eyes, width of eyes, height of hair, width of hair, style of hair, height of
nose, width of nose, width of ear and height of ear. The value for the height of
face was determined from the consumption of ice cream, the value for the
structure of face was determined from average family income and so on
repeating for the fifteen facial characteristics that define a Chernoff face.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("aplpack")#used to draw image
require("aplpack")
install.packages("Ecdat")# contains the data
require(Ecdat)

Step 2: Clean and prepare the data

data(Icecream )#load the date
#add the dates
dates<-c("18 Mar","17 Apr","16 May","14 Jun","13 Jul","11 Aug","09
Sep","08 Oct","06 Nov","05 Dec","03 Jan","01 Feb","01 Mar","30 Mar","28
Apr","27 May","25 Jun","24 Jul","22 Aug","20 Sep","19 Oct","17 Nov","16
Dec","14 Jan","12 Feb","13 Mar","11 Apr","10 May","08 Jun","11 Jul")

Step 3: Plot the chart

par(bg="cornsilk")# set background color to cornsilk

faces(Icecream,main="March 1952- July 1953",label=dates, face.type=1)

Visualization quick tips

Tip 1: To plot outline faces without color set face.type=0. You can also draw
Santa Claus faces by setting face.type=2.

Tip 2: Drop the argument labels if you don’t want to have dates above the faces.
The dates will be replaced with numbers.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("aplpack")
require("aplpack")
install.packages("Ecdat")
require(Ecdat)
data(Icecream )
dates<-c("18 Mar","17 Apr","16 May","14 Jun","13 Jul","11 Aug","09
Sep","08 Oct","06 Nov","05 Dec","03 Jan","01 Feb","01 Mar","30 Mar","28
Apr","27 May","25 Jun","24 Jul","22 Aug","20 Sep","19 Oct","17 Nov","16
Dec","14 Jan","12 Feb","13 Mar","11 Apr","10 May","08 Jun","11 Jul")
par(bg="cornsilk")
faces(Icecream,main="March 1952- July 1953",label=dates, face.type=1)
NOTES

Write your notes below…

 56. Textual analysis Chernoff face plot

Figure 56 1 State of the Union address using words to define Chernoff faces

How to read the chart

The image displays Chernoff faces calculated from the most frequent words in
President Obama’s 2011 and 2012 State of the Union address. The variables
used to define the face were height of face, width of face, structure of face,
height of mouth, width of mouth , degree of smiling, height of eyes, width of
eyes, height of hair, width of hair, style of hair, height of nose, width of nose,
width of ear and height of ear. The value for the height of face was determined
from the most common word (“America”) , the value for the structure of face
was determined by the next most popular word (“People”) and so on.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("aplpack")
require("aplpack")

Step 2: Clean and prepare the data

# Create variables with count of most popular words
c.2010<- c(70,33,29,29,20,13,12,11,13,15,13,15,6,11,10)
c.2011<-c(57,33,31,20,14,18,16,16,12,9,10,7,15,9,10)
words<-rbind(c.2010,c.2011)
#Add column names for most common words
colnames(words) <- c("America","People","Jobs","Business",
"Time","Government","Nation", "Tonight","Country","Energy",
"Taxes","Economy","Futurev","Care","Congress")
rownames(words)<-c("2011","2012")# add year 2011 and 2012

Step 3: Plot the chart

par(bg="papayawhip")# set the background color
faces(words,face.type=1,main="President Obama's State of the Union
Address",ncol.plot=1)

Visualization quick tips

Tip 1: To plot outline faces without color set face.type=0. You can also draw
Santa Claus faces by setting face.type=2.

Tip 2: Replace ncol.plot with nrow.plot if you wish to plot the faces as rows
rather than in a column.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("aplpack")
require("aplpack")
c.2010<- c(70,33,29,29,20,13,12,11,13,15,13,15,6,11,10)
c.2011<-c(57,33,31,20,14,18,16,16,12,9,10,7,15,9,10)
words<-rbind(c.2010,c.2011)
colnames(words) <- c("America","People","Jobs","Business",
"Time","Government","Nation", "Tonight","Country","Energy",
"Taxes","Economy","Futurev","Care","Congress")
rownames(words)<-c("2011","2012")
par(bg="papayawhip")
faces(words,face.type=1,main="President Obama's State of the Union
Address",ncol.plot=1)
NOTES

Write your notes below…

 57. Chernoff face time series plot

Figure 57 1 Ice cream consumption verses temperature

How to read the chart

The image displays a time series plot of ice cream consumption and temperature
using Chernoff faces. The Chernoff faces are calculated from four variables,
consumption of ice cream per head (in pints); average family income per week
(in US Dollars); price of ice cream (per pint) and the average temperature (in
Fahrenheit). The variables used to define the face were height of face, width of
face, structure of face, height of mouth, width of mouth , degree of smiling,
height of eyes, width of eyes, height of hair, width of hair, style of hair, height of
nose, width of nose, width of ear and height of ear. The value for the height of
face was determined from the consumption of ice cream, the value for the
structure of face was determined from average family income and so on
repeating for the fifteen facial characteristics that define a Chernoff face.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("aplpack") #package used to draw image
require("aplpack")
install.packages("Ecdat") # contains the data
require(Ecdat)

Step 2: Clean and prepare the data

data(Icecream )# data used in plot
b<- Icecream[1:6,]# We only want first 6 observations
#attach dates to the data
rownames(b)<-c <-c("18 Mar","17 Apr","16 May","14 Jun","13 Jul","11 Aug")
consumption= b[1:6,1] *100
temperature = b[1:6,4]

Step 3: Plot the chart

#first set up the basic plot area using x and y axis
plot(consumption, temperature,bty="n")
par(bg="lightblue")# set background color
b<-faces(b,plot=FALSE) # create face data frame
plot.faces(b, consumption, temperature,width=2, height=3)

Visualization quick tips

Tip 1: To plot outline faces without color set face.type=0. You can also draw
Santa Claus faces by setting face.type=2.

Tip 2: Adjust the width and height of the faces in the plot by using the argument
width and height. It may take some experimentation to get the size correct when
using your own data.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("aplpack")
require("aplpack")
install.packages("Ecdat")
require(Ecdat)
data(Icecream )
b<- Icecream[1:6,]
rownames(b)<-c <-c("18 Mar","17 Apr","16 May","14 Jun","13 Jul","11 Aug")
consumption= b[1:6,1] *100
temperature = b[1:6,4]
plot(consumption, temperature ,bty="n")
par(bg="lightblue")
b<-faces(b,plot=FALSE)
plot.faces(b, consumption, temperature,width=2, height=3)
NOTES

Write your notes below…

 58. Chernoff face rank order plot

Figure 58 1 Final position of top six and bottom six English football teams

How to- read the chart

The image displays the Chernoff faces of the top three and bottom three English
premier league football teams for the season 2012-2013. The Chernoff faces are
calculated from the following variables:

1 "height of face "= "Points"

2 "width of face " ="Goal Difference"
3 "structure of face"= "Home wins"
4 "height of mouth " = "Home for"
5 "width of mouth " = "Away wins"
6 "smiling " = "Away for"
7 "height of eyes "= "Home draw"
8 "width of eyes " = "Home loss"
9 "height of hair " = "Home against"
10 "width of hair " = "Away draw"
11 "style of hair " = "Away loss"
 12 "height of nose " = "Away against"
13 "width of nose " = "Points"
14 "width of ear " = "Goal Difference"
15 "height of ear " = "Home wins"

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("aplpack")# package used to draw plot
require("aplpack")

Step 2: Clean and prepare the data

#enter the data for each team
Man_Utd =c(89, 43, 16, 45,
12, 41, 0, 3, 19, 5, 2,
24)
Man_City=c( 78, 32, 14, 41,
9, 25, 3, 2, 15, 6, 4,
19)
Chelsea =c(75, 36, 12, 41, 10,
34, 5, 2, 16, 4, 5, 23)
Wigan=c( 36, -26, 4, 26, 5,
21, 6, 9, 39, 3, 11, 34)
Reading=c (28, -30, 4, 23, 2,
20, 8, 7, 33, 2, 15, 40)
QPR=c( 25, -30, 2, 13, 2,
17, 8, 9, 28, 5, 12, 32)
#combine teams together
names<-rbind(Man_Utd,Man_City,Chelsea,Wigan,Reading,QPR)
#add column names
colnames(names)<-c("Points", "Goal Difference", "Home
wins", "Home for", "Away wins", "Away for",
"Home draw", "Home loss", "Home against", "Away
draw", "Away loss", "Away against")

Step 3: Plot the chart

#convert names to a face object for plotting
teams<-faces(names,plot.faces=FALSE)
#set background color and color for text associated with faces
par(bg=" honeydew1",col="purple")

#create basic plot area

plot(0:6,0:6,type="n",axes=FALSE,xlab="Season 2012-2013 final league
position",ylab="",main="English Premier League")
#set up the custom axis
axis(side=1,at=(1:6), labels=c("1st","2nd","3rd","18th","19th","last"))
#plot the data
plot(teams,x.pos=1:6,y.pos=6:1)

Visualization quick tips

Tip 1: To plot outline faces without color set face.type=0. You can also draw
Santa Claus faces by setting face.type=2.

Tip 2: Adjust the width and height of the faces in the plot by using the argument
width and height. It may take some experimentation to get the size correct when
using your own data.

Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("aplpack")
require("aplpack")
Man_Utd =c(89, 43, 16, 45,
12, 41, 0, 3, 19, 5, 2,
24)
Man_City=c( 78, 32, 14, 41,
9, 25, 3, 2, 15, 6, 4,
19)
Chelsea =c(75, 36, 12, 41, 10,
34, 5, 2, 16, 4, 5, 23)
Wigan=c( 36, -26, 4, 26, 5,
21, 6, 9, 39, 3, 11, 34)
Reading=c (28, -30, 4, 23, 2,
20, 8, 7, 33, 2, 15, 40)
QPR=c( 25, -30, 2, 13, 2,
17, 8, 9, 28, 5, 12, 32)
names<-rbind(Man_Utd,Man_City,Chelsea,Wigan,Reading,QPR)
colnames(names)<-c("Points", "Goal Difference", "Home
wins", "Home for", "Away wins", "Away for",
"Home draw", "Home loss", "Home against", "Away
draw", "Away loss", "Away against")
teams<-faces(names,plot.faces=FALSE)
par(bg=" honeydew1",col="purple")
plot(0:6,0:6,type="n",axes=FALSE,xlab="Season 2012-2013 final league
position",ylab="",main="English Premier League")
axis(side=1,at=(1:6), labels=c("1st","2nd","3rd","18th","19th","last"))
plot(teams,x.pos=1:6,y.pos=6:1)
NOTES

Write your notes below…

 59. Constant feature Chernoff faces

Figure 59 1 Cricket teams playing in the 1890 County Championship

How to read the chart

The image displays constant Chernoff faces calculated from the points scored,
number of wins, losses and draws for teams playing in the first English County
Championship. The images are ranked from division winners (Surrey) to last
place (Sussex).The Chernoff faces are calculated from the following variables:

1 "height of face " = "Points"

2 "width of face "= "Wins"

3 "structure of face" = "Draws"

4 "height of mouth " = "Losses"

The remaining eleven Chernoff face features are given a constant value.

How to create this chart in less than ten minutes

Step 1: Download and install the required packages.

install.packages("aplpack")# package used to draw plot
require("aplpack")
Step 2: Clean and prepare the data
#enter the data for each team
Surrey= c ( 6, 9, 2, 3)
Lancashire= c( 4, 7, 4, 3)
Kent = c(3, 6, 5, 3)
Yorkshire= c ( 3, 6, 5, 3)
Nottinghamshire= c ( 0, 5, 4, 5)
Gloucestershire= c( -1, 5, 3, 6)
Middlesex=c ( -5, 3, 1, 8)
Sussex = c (-10, 1, 0, 11)
#combine teams together
names<-
rbind(Surrey,Lancashire,Kent,Yorkshire,Nottinghamshire,Gloucestershire,Middlesex,Sus
#add column names
colnames(names)<-c("Points", "Wins", "Draws",
"Losses")
#create a matrix of constants for the unchanging face features
constant <- matrix(1,nrow=nrow(names),ncol=11)
#combine data on points, wins, draws, losses with constant data
names<-cbind(names,constant)
Step 3: Plot the chart
par(bg=" honeydew1")#set background color
faces (names, main=" 1890 County Championship", face.type=1)

Visualization quick tips

Tip 1: To plot outline faces without color set face.type=0. You can also draw
Santa Claus faces by setting face.type=2.

Tip 2: Adjust the width and height of the faces in the plot by using the argument
width and height. It may take some experimentation to get the size correct when
using your own data.
Tip 3: Change the background color by setting the argument bg="your chosen
color".

Quick reference R code

install.packages("aplpack")
require("aplpack")
Surrey= c ( 6, 9, 2, 3)
Lancashire= c( 4, 7, 4, 3)
Kent = c(3, 6, 5, 3)
Yorkshire= c ( 3, 6, 5, 3)
Nottinghamshire= c ( 0, 5, 4, 5)
Gloucestershire= c( -1, 5, 3, 6)
Middlesex=c ( -5, 3, 1, 8)
Sussex = c (-10, 1, 0, 11)

names<-rbind( Surrey,Lancashire,Kent,Yorkshire,Nottinghamshire,
Gloucestershire,Middlesex,Sussex)
colnames(names)<-c("Points", "Wins", "Draws",
"Losses")
constant <- matrix(1,nrow=nrow(names),ncol=11)
names<-cbind(names,constant)
par(bg=" honeydew1")
faces (names, main=" 1890 County Championship", face.type=1)
NOTES

Write your notes below…

OTHER BOOKS YOU WILL ALSO ENJOY

100 Statistical Tests in R: 100 Statistical Tests in R is designed to give you

rapid access to one hundred of the most popular statistical tests. It shows you,
step by step, how to carry out these tests in the free and popular R statistical
package. The book was created for the applied researcher whose primary focus
is on their subject matter rather than mathematical lemmas or statistical theory.
Step by step examples of each test are clearly described, and can be typed
directly into R as printed on the page. To accelerate your research ideas, over
three hundred applications of statistical tests across engineering, science, and
the social sciences are discussed.

www.auscov.com
[i] This dataset is rather famous and known as Anscombe's quartet. See Anscombe, F. J. (1973). "Graphs in
Statistical Analysis". American Statistician 27 (1): 17–21. JSTOR 2682899.

[ii] See page 206 in Galton, Francis. Natural inheritance. Vol. 42. MacmillaSn, 1889.
[iii] The height of the boxes are proportional to the Pearson residuals with the width proportional to the
square roots of the expected frequencies.

[iv] The crime data consisted of the following crimes: Murder, Rape Robbery,
Assault, Burglary, Larceny and Motor Theft.

[v] See http://dictyexpress.biolab.si/index.htm

[vi] See www. http://newsmap.jp/#/b,e,m,n,s,t,w/us/view/

[vii] See Gao X, An H, Zhong W (2013) Features of the Correlation Structure of Price Indices. PLoS ONE
8(4): e61091. doi:10.1371/journal.pone.0061091

[viii] Image taken from the website http://www.florence-nightingale-avenging-angel.co.uk/ . Also see the
insightful book by Hugh Small entitled Florence Nightingale, Avenging Angel.