Summary Statistics

Visualization

Descriptive Statistics

Felipe José Bravo Márquez

August 26, 2021

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Exploratory Data Analysis

Descriptive Statistics or Exploratory Data Analysis or (EDA) encompasses a set

of techniques to quickly understand the nature of a data collection or dataset.
The main goal of descriptive statistics is to explore the data to find some patterns
that can be exploited to generate hypotheses.
It was proposed by the statistician John Tukey.
It is based mainly on two types of techniques: summary statistics and data
visualization.
In this class you will see both types of techniques, in addition to their application
in R for some toy datasets.
This class in partially based on chapter 3 of [Tan et al., 2016].

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

The Iris dataset

We will work with a well-known dataset called Iris.
The dataset consists of 150 observations of iris plant flowers.
There are three types of iris flower classes: virginica, setosa and versicolor.
There are 50 observations of each.
The variables or attributes measured for each flower are:
1 The type of flower as a categorical variable.
2 The length and width of the petal in cm as numerical variables.
3 The length and width of the sepal in cm as numeric variables.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

The Iris dataset

Figure: Virginica - Setosa - Versicolor

The dataset is available in R:

>data(iris) # to load the dataset into the workspace
> names(iris)
[1] "Sepal.Length" "Sepal.Width" "Petal.Length"
"Petal.Width" "Species"

In order to access the variables directly in our workspace we use the command
attach(iris).

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Summary Statistics

Summary statistics are values that explain properties of the data.

Some of these properties include: frequencies, measures of central tendency
and dispersion.
Example:
Central tendency: mean, median, mode.
Variation: measure the variability of the data, such as standard deviation,
range, etc..
Most summary statistics can be calculated by making a single pass through the
data.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Frequency and Mode

The absolute frequency of an attribute value is the number of times it is
observed.
The relative frequency is the absolute frequency divided by the total number of
examples.
In R we can count the frequencies of occurrence of each distinct value of a
vector using the command table:
> table(iris$Species)
setosa versicolor virginica
50 50 50
> vec<-c(1,1,1,0,0,3,3,3,3,2)
> table(vec)
vec
0 1 2 3
2 3 1 4

Exercise: Calculate the relative frequencies of the above vector.

> table(vec)/length(vec) # Relative frequencies
vec
0 1 2 3
0.2 0.3 0.1 0.4

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Frequency and Mode (2)

The mode of an attribute is the most frequent value observed.

The mode function is not implemented natively in R, but it is easy to calculate
using table and max:
my_mode<-function(var){
frec.var<-table(var)
value<-which(frec.var==max(frec.var))
as.numeric(names(value))
}
> my_mode(vec)
[1] 3
> my_mode(iris$Sepal.Length)
[1] 5

We generally use frequencies and mode to study categorical variables.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Central Tendency Measures

These measures attempt to summarize the observed values into a single value
associated with the centrally located value.
The mean is the most common measure of central tendency for a numeric
variable.
If we have n observations it is calculated as the arithmetic mean or average.
n
1X
mean(x) = x = xi
n
i=1

The major problem with the mean is that it is very sensitive to outliers.
We take a random vector of mean 20 and then add a random element that
comes from a distribution of much larger mean. We see that the mean is strongly
affected by noise:
> vec<-rnorm(10,20,10)
> mean(vec)
[1] 16.80036
> vec.noise<-c(vec,rnorm(1,300,100))
> mean(vec.noise)
[1] 35.36422

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Central Tendency Measures (2)

We can robust the mean by removing a fraction of the extreme values using the
trimmed mean.
In R we can give a second parameter to the function mean called trim that
defines the fraction of extreme elements to discard.
Example: We discard 10% of the extreme values in the previous example:
> mean(vec,trim=0.1)
[1] 17.78799
> mean(vec.noise,trim=0.1)
[1] 19.51609 # much more robust

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Central Tendency Measures (3)

The median represents the central ranking position of the variable that separates
the lower half and the upper half of the observations.
Intuitively, it consists of the value where for one half of the observations all values
are greater than it, and for the other half all are less.

xr +1 If |x| (vector length) is odd, |x| = 2r + 1
median(x) = 1
(x
2 r
+ xr +1 ) If |x| is even, |x| = 2r

For the above example, we see that the median is more robust to noise than the
mean:
> median(vec)
[1] 17.64805
> median(vec.noise)
[1] 17.64839

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Comparison between mode, median and mean

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Percentiles or Quantiles

The k-th percentile of a numerical variable is a value such that k% of the

observations are below the percentile and (100 − k)% are above this value.
Quantiles are equivalent to percentiles but expressed in fractions instead of
percentages.
In R they are calculated with the command quantile:
# All percentiles
quantile(Sepal.Length,seq(0,1,0.01))

In addition it is very common to talk about the quartiles which are three specific
percentiles:
The first quartile Q1 (lower quartile) is the percentile with k = 25.
The second quartile Q2 is with k = 50 which is equivalent to the median.
The third quartile Q3 (upper quartile) is with k = 75.
# The minimum, the three quartiles and the maximum.
> quantile(Sepal.Length,seq(0,1,0.25))
0% 25% 50% 75% 100%
4.3 5.1 5.8 6.4 7.9

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Summarizing a Data Frame

In R we can summarize various summary statistics of a variable or of a
data.frame using the command summary.
For numerical variables it gives us the minimum, the quartiles, the mean and the
maximum.
For categorical variables it gives us the frequency table.

> summary(iris)
Sepal.Length Sepal.Width Petal.Length Petal.Width
Min. :4.300 Min. :2.000 Min. :1.000 Min. :0.100
1st Qu.:5.100 1st Qu.:2.800 1st Qu.:1.600 1st Qu.:0.300
Median :5.800 Median :3.000 Median :4.350 Median :1.300
Mean :5.843 Mean :3.057 Mean :3.758 Mean :1.199
3rd Qu.:6.400 3rd Qu.:3.300 3rd Qu.:5.100 3rd Qu.:1.800
Max. :7.900 Max. :4.400 Max. :6.900 Max. :2.500

Species
setosa :50
versicolor:50
virginica :50

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Exercise

Using the command tapply analyze the mean, median and quartiles for the
three species of Iris for the four variables.
Do you notice any differences in the different species?

tapply(iris$Petal.Length,iris$Species,summary)
tapply(iris$Petal.Width,iris$Species,summary)
tapply(iris$Sepal.Length,iris$Species,summary)
tapply(iris$Sepal.Width,iris$Species,summary)

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Variability Measures
Variability mesaures or dispersion measures tell us how different or similar the
observations tend to be with respect to a particular value. Usually this value
refers to some measure of central tendency.
The range is the difference between the maximum and minimum value:
> max(Sepal.Length)-min(Sepal.Length)
[1] 3.6

The standard deviation is the square root of the variance that measures the
mean squared differences of the observations from the mean.

n
1 X
var(x) = (xi − x)2
n−1
i=1

p
sd(x) = var(x)
> var(Sepal.Length)
[1] 0.6856935
> sd(Sepal.Length)
[1] 0.8280661

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Variability Measures (2)

Like the mean, the standard deviation is sensitive to outliers.
The most robust measures are usually based on the median.
Let m(x) be a measure of central tendency of x (usually the median), we define
the average absolute deviation (AAD) as:

n
1X
AAD(x) = |xi − m(x)|
n
i=1

Exercise: Program the function add in R, as a function that receives a vector x

and a central mean function fun. The absolute value is calculated with the
command abs:
aad<-function(x,fun=median){
mean(abs(x-fun(x)))
}
> aad(Sepal.Length)
[1] 0.6846667
> aad(Sepal.Length,mean)
[1] 0.6875556

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Variability Measures (3)

Let b be a constant we define the median absolute deviation as:

MAD(x) = b × median(|xi − m(x)|)

In R is calculated with the command mad with the parameters center as a

function measuring the central tendency of the variable and constant as the
constant b. By default the median and the value 1.482 is used.
> mad(Sepal.Length)
[1] 0.7

Finally, the interquartile range (IQR) is defined as the difference between the
third and the first quartile (Q3 − Q1 ).
IQR(Sepal.Length)
[1] 1.3

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Multivariate Summary Statistics

To compare how one variable varies with respect to another, we use multivariate
measures.
The covariance cov (x, y) measures the degree of joint linear variation of a pair
of variables x, y:
n
1 X
cov (x, y) = (x − x)(y − y)
n−1
i=1

Where cov (x, x) = var (x)

In R it is computed with the command cov:
> cov(Sepal.Length,Sepal.Width)
[1] -0.042434
If we give it a matrix or a data.frame of numeric variables, it computes a
covariance matrix:
> cov(iris[,1:4])
Sepal.Length Sepal.Width Petal.Length Petal.Width
Sepal.Length 0.6856935 -0.0424340 1.2743154 0.5162707
Sepal.Width -0.0424340 0.1899794 -0.3296564 -0.1216394
Petal.Length 1.2743154 -0.3296564 3.1162779 1.2956094
Petal.Width 0.5162707 -0.1216394 1.2956094 0.5810063

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Multivariate Summary Statistics (2)

If two variables are independent of each other, their covariance is zero.

A measure of relationship that does not depend on the scale of each variable is
the linear correlation.
The linear correlation or Pearson’s correlation coefficient r (x, y) is defined as:

cov (x, y)
r (x, y ) =
sd(x)sd(y)

The linear correlation varies between −1 to 1.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Multivariate Summary Statistics (3)

A value close to 1 indicates that as one variable grows the other also grows in a
linear proportion.
A value close to -1 indicates an inverse relationship (one is growing and the
other is decreasing).
If the correlation is close to zero we have linear independence.
Note that a correlation of zero does not imply that there cannot be a non-linear
relationship between the variables.
In R, the correlation is calculated with the command cor.
> cor(iris[,1:4])
Sepal.Length Sepal.Width Petal.Length Petal.Width
Sepal.Length 1.0000000 -0.1175698 0.8717538 0.8179411
Sepal.Width -0.1175698 1.0000000 -0.4284401 -0.3661259
Petal.Length 0.8717538 -0.4284401 1.0000000 0.9628654
Petal.Width 0.8179411 -0.3661259 0.9628654 1.0000000

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Contingency Tables

To analyze the relationship between categorical variables we use contingency

tables.
The table is filled with the marginal frequencies of all pairs of values between two
categorical variables.
In R they are created using the command table that we used before for
computing frequencies, but now giving two vectors as input:
gender<-c("Male", "Female", "Male", "Female", "Female", "Male")
studies<-c("college","postgraduate","high school",
"postgraduate","high school","college")

>table(gender,studies)
studies
gender college high school postgraduate
Female 0 1 2
Male 2 1 0

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Skweness and Kurtosis

There are other two summary statistics that focus on more complex properties of the
data distribution called skeweness and kurtosis [Pipis, 2020].

Skeweness
Skeweness is a measure of the asymmetry of the probability distribution.
Pn 3
i=1 (xi − x)
skewness(x) =
(n − 1)sd(x)3

It indicates how much our underlying distribution deviates from the normal
distribution since the normal distribution has skewness 0.
Generally, we have three types of skewness:
1 Symmetrical: the skewness is close to 0 and the mean is almost the
same as the median.
2 Negative skew: the majority of the observations are concentrated on the
right tail (the median is greater than the mean).
3 Positive skew: the majority of the observations are concentrated on the
left tail (the median is less than the mean).

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Skweness

Figure: Source: Wikipedia

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Skweness

In R, we can calculate skweness with the formula skewness from the library
moments:
> library(moments)
#positive skew
> skewness(c(1,1,2,3))
[1] 0.4933822
#symmetrical
> skewness(c(1,2,3))
[1] 0
#negative skew
> skewness(c(1,2,3,3))
[1] -0.4933822

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Kurtosis

Kurtosis
The Kurtosis describes the “tailedness” of a distribution [Westfall, 2014].
Pn 4
i=1 (xi − x)
kurtosis(x) =
(n − 1)sd(x)4
Let’s see the main three types of kurtosis.
1 Mesokurtic: This is the normal distribution (kurtosis ≈ 3)
2 Leptokurtic: This distribution has fatter tails than a normal distribution
(kurtosis > 3). Consequently, outliers are more likely to occur than in a
normal distribution.
3 Platykurtic: The distribution has thinner tails than a normal distribution
(kurtosis < 3). Consequently, outliers are less likely to occur than in a
normal distribution.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Kurtosis

Figure: Source: tutorialspoints

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Kurtosis
In R, we can calculate kurtosis with the formula kurtosis from the library
moments.
Let’s calculate kurtosis for normal data:
> x <- rnorm(1000, 0,1)
> plot(density(x))
> kurtosis(x)
[1] 2.946869

As expected we got a value close to 3.

density.default(x = x)
0.3
Density

0.2
0.1
0.0

−4 −2 0 2 4

N = 1000 Bandwidth = 0.2303

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Kurtosis
Now using random data generated with an exponential long-tailed distribution:
> x<-rexp(1000)
> plot(density(x))
> kurtosis(x)
[1] 6.839485

As expected we get a positive excess kurtosis (i.e. greater than 3) since the
distribution has fatter tails (Leptokurtic).

density.default(x = x)
0.6
Density

0.4
0.2
0.0

0 2 4 6 8

N = 1000 Bandwidth = 0.1916

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Kurtosis
Now using data generated with a thinner-tailed Beta distribution with
hyperparameters 2, 2:
> x <-rbeta(1000,2,2)
> plot(density(x))
> kurtosis(x)
[1] 2.1046

As expected we get a negative excess kurtosis (i.e. less than 3) since the
distribution has thinner tails (Platykurtic).

density.default(x = x)
1.5
1.0
Density

0.5
0.0

0.0 0.2 0.4 0.6 0.8 1.0

N = 1000 Bandwidth = 0.05163

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Data Visualization

Data visualization is the transformation of a dataset into a visual format that

allows people to identify the characteristics and relationships between examples
in the dataset.
Visualization allows people to recognize patterns or trends based on their
judgment or expertise in the particular domain.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Representation

Representation is understood as the mapping of data into a visual format

Examples, their attributes and relationships are translated into graphical
elements such as points, lines, shapes and colors.
Examples are usually represented as points.
Attribute values are represented as the position of the points or the
characteristics of the points, e.g. color, size and shape.
When using position to represent values it is simple to detect if groups of objects
are formed or the presence of outliers.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Plotting in R

In R the most frequent display function is plot.

t is a generic function whose result depends on the nature of the variables given
as input.
To all plots we can add additional parameters such as: main for the title, xlab
and ylab for the name of the x-axis and y-axis.
Other properties are: col to define the color, type to define the chart type: (p)
for points or (l) for lines.
We can also add new layers to a plot with the command lines.
To save an image to a file we can use Rstudio’s export button.
To do it from the R command line:
png("image.png")
plot(1:10)
dev.off()

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Example
plot(rnorm(15,10,5),col="red",type="p",pch=1)
lines(rnorm(15,10,5),col="blue",type="p",pch=1)
lines(rnorm(15,10,5),col="green",type="b",pch=2)
title(main="My Plot")
legend(’topright’, c("lines","dots","both") ,
lty=1:3, col=c("red", "blue","green"), bty=’n’, cex=.75)

My Plot
20

lines
dots
both
15
rnorm(15, 10, 5)

10
5
0

2 4 6 8 10 12 14

Index

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Histograms
They show the distribution of the values of a variable.
The values of the elements are divided into bins and bar charts are created for
each of them.
The height of each bar indicates the number of examples in the corresponding
bin.
In R they are created with the command hist.
> hist(Sepal.Length)

Histogram of Sepal.Length
Frequency
25
10
0

4 5 6 7 8
Sepal.Length
Felipe Bravo Márquez Descriptive Statistics
 Summary Statistics
Visualization

Histograms (2)

The shape of the histogram depends on the number of bins.

In R this number can be defined with the parameter nclass.
> hist(Sepal.Length,nclass=100)

Histogram of Sepal.Length
Frequency
8
4
0

4.5 5.5 6.5 7.5

Sepal.Length

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Histograms (3)
A very popular library for making visualizations in R, which is part of tidyverse, is
ggplot2.
It is based on the idea of decomposing the plot into semantic components such
as scales and layers.
>install.packages("ggplot2")
>library(ggplot2)
>ggplot(iris, aes(x=Sepal.Length))
+ geom_histogram(bins = 10, color="black", fill="white")

20
count

10

5 6 7 8
Sepal.Length

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Density
Another way to visualize how the data are distributed is to estimate a density.
These are calculated using nonparametric statistical techniques called kernel
density estimation.
The density is a smoothed version of the histogram and allows us to determine
more clearly if the observed data behaves like a known density e.g., Gaussian.
In R they are created with the command density, and then visualized with the
command plot.
plot(density(iris$Sepal.Length),main=""Density of Sepal.Length")

Density of Sepal.Length
0.4
0.3
Density

0.2
0.1
0.0

4 5 6 7 8

N = 150 Bandwidth = 0.2736

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Pie Charts

The pie charts represent the frequency of the elements in a circle.

Each category has a share proportional to its relative frequency.
They are generally used for categorical variables:
pie(table(iris$Species))

setosa

versicolor

virginica
Many people consider pie charts to be misleading and recommend using
histograms as a better alternative [Poldrack, 2019].

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots

Boxplots are built from the percentiles.

A rectangle is constructed between the first and third quartiles (Q1 and Q3 ).
The height of the rectangle is the interquartile range IQR (Q3 − Q1 ).
The median is a line that divides the rectangle.
Each end of the rectangle is extended with a line or arms of length
Q1 − 1.5∗IQR for the lower line and Q3 + 1.5∗IQR for the upper line.
Values more extreme than the length of the arms are considered outliers.
The boxplot gives us information about the symmetry of the data distribution.
If the median is not in the center of the rectangle, the distribution is not
symmetrical.
They are useful to detect the presence of outliers.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (2)

Figure: https://www.laboneconsultoria.com.br/
wp-content/uploads/2018/07/Boxplot-04.png

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (3)
The length of the arms as well as the criteria for identifying outliers is based on
the behavior of a normal distribution.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (4)

In R boxplots are plotted with the command boxplot:

> boxplot(Sepal.Length,main="Boxplot Sepal.Length")

Boxplot Sepal.Length
7.5
6.0
4.5

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (4)

If we have a factor variable we can create a boxplot for each category as follows:
> boxplot(Sepal.Length˜Species,ylab="Sepal.Length")

7.5
Sepal.Length
6.5
5.5
4.5

setosa versicolor virginica

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (5)

We can also compare several boxplots in the same plot:

> boxplot(x=iris[,1:4],main="Boxplots Iris")

Boxplots Iris
8
6
4
2
0

Sepal.Length Sepal.Width Petal.Length Petal.Width

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Boxplots (6)
Now using ggplot2:
> ggplot(iris, aes(x = Species, y = Sepal.Length,
fill = Species)) + geom_boxplot()
8

Species
Sepal.Length

setosa

6 versicolor
virginica

setosa versicolor virginica

Species

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots

Scatter plots use Cartesian coordinates to display the values of two numerical
variables of the same length.
The values of the attributes determine the position of the elements.
Other attributes can be used to define the size, shape or color of objects.
In R we can plot a scatterplot of two numeric variables using the command
plot(x,y), which would be y vs x.
We can also define formulas f (x) = y using the notation y˜x.
Thus, the command plot(y˜x) is equivalent to plot(x,y).
If we have a data.frame or a numerical matrix, we can see the scatterplots of all
pairs using the command pairs(x).

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots (2)

Examples:
# Sepal width vs. sepal length
plot(Sepal.Width˜Sepal.Length, col=Species)
# Equivalent
plot(Sepal.Length, Sepal.Width,col=Species,
pch=as.numeric(Species))
# We add a legend
legend(’topright’, levels(Species) ,
lty=1, col=1:3, bty=’n’, cex=.75)

setosa
4.0

versicolor
Sepal.Width

virginica
3.0
2.0

4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

Sepal.Length

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots (3)

The same using ggplot2:
ggplot(iris, aes(x=Sepal.Length,
y=Sepal.Width, color=Species)) +
geom_point(size=3,shape=4)

4.5

4.0

3.5
Species
Sepal.Width

setosa
versicolor

3.0 virginica

2.5

2.0

5 6 7 8
Sepal.Length

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots (4)

All pairs of the 4 variables of the iris dataset using a different color and character
for each species:
pairs(iris[,1:4],pch=as.numeric(iris$Species),col=iris$Species)

2.0 3.0 4.0 0.5 1.5 2.5

4.5 6.0 7.5

Sepal.Length
2.0 3.0 4.0

Sepal.Width

7
5
Petal.Length

3
1
0.5 1.5 2.5

Petal.Width

4.5 5.5 6.5 7.5 1 2 3 4 5 6 7

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots (5)

Three-dimensional scatterplots can also be created.

The scatterplot3d library must be installed using the following command:
install.packages("scatterplot3d",dependencies=T)

Then load the library by typing library(scatterplot3d).

A 3d scatterplot for petal width, sepal length and sepal width:
scatterplot3d(iris$Petal.Width, iris$Sepal.Length,
iris$Sepal.Width, color=as.numeric(iris$Species),
pch=as.numeric(iris$Species))

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Scatter Plots (6)

4.5
4.0
Sepal.Width
3.5

Sepal.Length
3.0

8
7
2.5

6
5
2.0

4
0.0 0.5 1.0 1.5 2.0 2.5

Petal.Width

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Parallel Coordinate Plots

Parallel coordinate plots are another way to visualize multi-dimensional data.

Instead of using perpendicular axes (x-y-z) we use several axes parallel to each
other.
Each attribute is represented by one of the parallel axes with its respective
values.
The values of the different attributes are scaled so that each axis has the same
height.
Each observation represents a line that joins the different axes according to their
values.
In this way, examples similar to each other tend to be grouped in lines with
similar trajectory.
In many occasions it is necessary to re-order the axes in order to visualize a
pattern.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Parallel Coordinate Plots (2)

In R we can create parallel coordinate graphs with the command parcoord
from the MASS library.
Example:
library(MASS)
parcoord(iris[1:4], col=iris$Species,var.label=T)

7.9 4.4 6.9 2.5

4.3 2.0 1.0 0.1

Sepal.Length Sepal.Width Petal.Length Petal.Width

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

Conclusions

We have learned how to describe a dataset using various metrics and

visualizations.
These techniques give us a general understanding of the data.

Felipe Bravo Márquez Descriptive Statistics

 Summary Statistics
Visualization

References I

Pipis, G. (2020).
Skewness and kurtosis in statistics: R-bloggers.
Poldrack, R. A. (2019).
Statistical thinking for the 21st century.
https://statsthinking21.org/.
Tan, P.-N., Steinbach, M., and Kumar, V. (2016).
Introduction to data mining.
Pearson Education India.
Westfall, P. H. (2014).
Kurtosis as peakedness, 1905–2014. rip.
The American Statistician, 68(3):191–195.

Felipe Bravo Márquez Descriptive Statistics