Scribd
Upload
15 views

08 - Practice of Data Visualization

The document discusses best practices for data visualization, including box plots, stacked area plots, cartograms, color scales, and Tufte's principles of maximizing data ink ratio, minimizing chartjunk, and proper scaling. Examples of effective visualizations include John Snow's cholera map and election state visualizations.

Uploaded by

natolihunduma11
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
15 views

08 - Practice of Data Visualization

The document discusses best practices for data visualization, including box plots, stacked area plots, cartograms, color scales, and Tufte's principles of maximizing data ink ratio, minimizing chartjunk, and proper scaling. Examples of effective visualizations include John Snow's cholera map and election state visualizations.

Uploaded by

natolihunduma11
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 44

Data Science

Steven Skiena
Stony Brook University

Practice of Data Visualization


Box and Whisker Plots
Box plots concisely show the range / quartiles
(i.e. median and variance) of a distribution.

I personally prefer contour


lines without the boxes.
Stacked Area vs. Line Plots
Hard to see trends in
middle areas of stack:
Stacked Bar Charts: Titanic
Data Maps and Cartograms
Cartograms distort regions to reflect an
underlying variable.
Non-Geographic Data Maps
Tools for Data Visualization
Just because Excel is very popular does not
mean it produces good graphs/plots.
The statistical language R has a very extensive
library of data visualizations.
MatPlotLib is your key to producing good
graphs/plots in Python.
Repetitions for Multivariate Data
Small multiple plots /
tables are good ways to
represent multivariate
data.
Understanding Color Scales
Rainbow Color Maps
Rainbows are perceptually non-linear.
Distinct positive/negative colors reflected about
a center make good scales.
Which Coloring is Better?
Tufte’s Visualization Aesthetic
Distinguishing good/bad visualizations requires
a design aesthetic, and a vocabulary to talk
about data representations:
● Maximize data ink-ratio
● Minimize lie factor
● Minimize chartjunk
● Use proper scales and clear labeling
Great Data Visualizations
● Display data accurately and clearly.
● Tell a story that the data reveals.
● Are rich enough to make you want to look
carefully and study the data.
Napoleon’s Advance and Retreat
New York’s Weather Year in Review
A clear story displaying over 3,000 numbers.
Marey’s Train Schedule
What can you see
here you cannot
with normal train
schedules?

It would be even
better with a
lighter datagrid.
Never imprison
your data!
Discovering the Source of Cholera
John Snow used this
data map to identify
the source of an 1854
Cholera epidemic as a
single contaminated
water pump.
Trying to Stop the Challenge Launch
Engineers failed to convince management to call off the
launch using a poor data visualization.
State of the Race
Effective use of what could
have been chartjunk and color
to capture the state of the
2020 election in one view
Which Executives Earn their Pay?
Chart plots points in 4D,
shown by x, y, color, size.
Using attributes like point
size/color is better than
plotting points in 3D.
Terrible Student Visualizations...
How do we fix this plot of word frequency?
Power Laws Need Log Plots!
Log-Log plots can be even more revealing.
Terrible Professional Visualizations..
● Display as little information as possible.
● Obscure your data with chart junk, like
pseudo-3D and excess color.
● Use poorly chosen scales.
Examples taken from http://wtfviz.net/
What is Wrong with this Pie Chart?
What’s Wrong with this Bar Chart?
Color and Dimensionality
Volume/Value Comparisons
Oval Pie Charts?
Range, Caption, and Symbol Sins
The Virtues of Consistency
Provably Meaningless Charts
Dramatic Misuse/Reuse of Color
Graphics Size Matters
Impressive Chart Junk
Labels Matter
Way Too Many D
Size and Ordering Implications
Keep a Critical Eye
Remember Tufte’s principles whenever
designing or interpreting data visualizations:
● Maximize data-ink ratio
● Minimize lie factor
● Minimize chartjunk
● Use proper scales and clear labeling
Beautiful data deserves beautiful visualization.
Term of the Week

You might also like