# Visualizing Data

Jeff Arnold
April 9, 2013 Emory University, Atlanta

Data Viz is Everywhere
Business / Economics Weather Sports Finance

Outline
Examples 1.  Florence Nightingale 2.  Challenger Explosion What is it? How does it work? When doesn't it work?

Examples

Florence Nightingale

Challenger Explosion

What is Data Visualization?

Grammar of Graphics

Grammar of Graphics

Grammar of Graphics
Geometric Shapes points lines bars text Aesthetics: convey information x position y position size of elements shape of elements color of elements

Data and Aesthetics

How does it work?
PATTERNS PATTERNS PATTRESN PATTERNS

Anscombe Quartet

Looking for Patterns

Plots are Comparisons
Actual Data Expected Data

When (and why) does it not work?
1.  2.  3.  4.  Too many variables Too many observations Perceptual biases Understanding randomness

Too Many Variables

Too Many Observations (I)

Too Many Observations (II)

Too Many Observations (III)

Visual Perception Biases
Q: What is the value of a ­ b? Does it change?

Visual Perception Biases
A: a ­ b = 2 everywhere.

Visual Perception Biases

Visual Perception Biases

Understanding Randomness
Q: In which plot were the points selected from a uniform random distribution?

Understanding Randomness
A: The plot on the right.

Conclusion
Data visualization and statistics are complementary Data visualization intuitive cognitive biases Statistical methods un­intuitive overcome our cognitive biases

Questions?

References
Nightingale receiving the Wounded at Scutari, By Jerry Barrett Diagram of the Causes of Mortality in the Army in the East, by Florence Nightingale Space Shuttle Challenger explodes shortly after take­off. Plot of GE vs. SP500 from Yahoo! Finance Kimmo Soramaki, Morten L. Bech, Jeffrey Arnold, Robert J. Glass and Walter E. Beyeler (2007). "The Topology of Interbank Payment Flows", Physica A.  url url

Hadley Wickham (2010). "A Layered Grammar of Graphics", Journal of Computational and Graphical Statistics.