How not to lie with visualization

1
How (not) to lie with visualization

• cs5984 Information Visualization
• Chris North

2
Final Quiz

• 7 information types
• 1d
• 2d
• 3d
• Multi-d
• Trees
• Graphs
• Doc collections

3
Where are we?

• Information Types
• Multi-D
• 1D
• 2D
• Hierarchies/Trees
• Networks/Graphs
• Document collections
• 3D

• Topics
• Design Principles
• Empirical Evaluation
• Java Development
• Visual Overviews
• Multiple Views
• Peripheral Views
• Workspaces
• Debates
• Vis. Lies

4
How (not) to lie with visualization

• Show and tell
• USA Today graphs

5
Stock Market Crash?!
Market
9000 8875 8750 8625 8500
1995 1996 1997 1998 1999 2000
6
Show entire scale
Market
10,000 7500 5000 2500 0
1995 1996 1997 1998 1999 2000
7
Show in context
Market
10,000 7500 5000 2500 0
1950 1960 1970 1980 1990 2000
8
Another example
9
Percentages 0 100
Employment rate 100 unemployment rate
10
Tuftes Rule

• Visual attribute value should be directly
  proportional to data attribute value
• Lie factor (visual effect) / (data effect)
• truth 1.0

11
Company financial status
12
The hidden 0-points

• Lie factor ?

13
Changing Scale
0.5?
13
14
ChangingScale
15
with linear time scale
16
Down Bad ?
17
Make it explicit
? Better
Other examples user performance, questionnaire
results
18
Logarithmic data ? log scale
19
Size Coding
20
Size Coding width or area?

?
21
Size Coding
22
Width or Area

• Width valueHeight valueArea value2
• or
• Area valuewidthheight valuewidth height
  value 0.5

Problem Using 2 dimensions to represent 1
dimension.
23
Volume coding?
Height? Diameter? Surface area? Volume?
73 79 data difference 5.5x 73 79 volume
difference 270x
24
Problem with area encoding
1 2 3 4 5 6 7
Width Area Volume
25
Width height encoding
1 2 3 4 5 6 7
Width Width Height
26
Solution just use width (or height)
27
A Propaganda Classic
28
Hmmm

• Low rank good!
• Different time scales
• Not really tuition
• Artistic mood

29
How not to lie

• Show entire scale
• Show data in context
• Consistent, linear scale
• Log scale for log data
• Up vs. down indicate direction of improvement
• Avoid size coding
• Use width OR height
• Dont use both for same data attribute
• Avoid area coding

30
Visualization Communication

• Communication is person dependent
• People have a lot of baggage

31
Expectations
Paris in the the spring
Lifeis aa highway
Nowis thethe time
32
Re-training

• Red spades, black hearts
• Poor user performance even after being told

33
Orientation

• Who are they?

34
Orientation
35
Homework 3 Results

• Hinite bites.
• Too much hypertexty stuff
• Not enough zooming, infovisy stuff
• Keep trying to break out of the box!

36
Projects

• Visualization in the periphery - evaluation
• David, Christa, Ali, Jon
• Visualizing Multi-D functions
• Reenal, Priya, Mrinmayee
• Visualization of data structures evaluation
• Kunal, Vikrant, Anuj
• Snap-together visualization
• Rohit, Varun, Jeevak, Atul
• Visual Break-down analysis with Financial data
• Ganesh, Anusha, Muthukumar, Sandeep
• Human-eye view
• Alex, Qiang, Ming, Vishal, Ahmed
• Bioinformatics
• Quoc, Mudita, Dhananjay
• Online chat/video-conference visualization
  (virtual school)
• Mahesh, Ben, Samal, Kuljeet, Harsha, Parool
• Digital libraries
• Jun, Supriya, Abhishek, Anil
• Maps and in-vehicle interfaces

Dec 4
Dec 6
Dec 11
37
Your Last InfoVis Assignment!

• Dec 18 Project Final Paper due
• Dec 7 ACM CHI short papers due
• Other destinations
• March? IEEE InfoVis papers due