Mapping
Mapping is a function to take data space to a visual space.
Data space
Datasets are formed of dimensions, and each dimension has a type.
The type informs effective mapping choices.
- Categorical: cats, dogs, birds
- Ordinal: small, medium, large
- Continuous: 1, 2, 3
Visual space
AKA "Marks"
There are many:
- position
- color
- shape
- orientation
- size
- value/lightness
- motion
- texture
Mapping strategy
What visual encoding do we use?
How do we compute the visual representation from the data value?
Expressiveness: Express all the data, but nothing else
Effectiveness: The importance of the attribute should match the salience of the channel
Tips
Log scales can help with uneven distributions of data, but can make your chart harder to read.
Mapping data to size can be misleading. EX: twice as big diameter makes a circle 4 times as large in terms of area.
QTons: use shape complexity to represent a gradient of change. Texture density helps show complexity increase.
Use motion very sparingly: direction, velocity, acceleration.
Texture: Density, variation, pattern
Color: Hue, chroma, luminance
Percentages should add up to 100%
Don't truncate the Y axis in general, if you think you need to perhaps your chart should graph the rate of change or something else
Data in 3d can lead to occlusion, it's better to use mapping tools to show extra dimensions
Avoid desert fog: zooming in should show where you are in the big picture. Overview + detail and focus+context
Detail
- Overview first
- Zoom and filter
- Details on demand
Interaction
Selection
- highlight
- low light
- filter
Explore
- panning
Abstract/elaborate
- zoom
- semantic zoom: break apart aggregates
filtering
- direct manipulation
- dynamic queries
encode
- change mapping
- layering visualization
coordination
- brushing
- linking