Graphical Perception
Principles
Goal: Communicate information
Author (you) -> graphic -> Reader
A graph is more effective than another if its quantitative information can be decoded more quickly or more easily by most observers.
Nancy Robbins, Creating More Effective Graphs
Detection: recognition of geometry encoding values
Assembly: grouping of detected symbol elements
Estimation: assessment of relative magnitude of values
Position along a common scale (most accurate task)
Position along identical nonaligned scales
Length
Angle and slope
Area
Volume
Color: hue (red, green, blue, etc.), saturation (pale/deep), and lightness. Hue can give good discrimination but poor ordering.
Show the data
Avoid distorting the data
Aid comparisons
Maximise data:ink
Label data, minimize legends
Use clearly different symbols, colours
Consider 0
Dot plots are often better than bar charts
Bar charts:
High ink:data ratio
Error bars cause perception errors
Can only show one-sided confidence intervals well
Thick bars reduce the number of categories that can be shown
Labels on vertical bar charts are difficult to read