Graphical Perception (2/2)

Outline

  • Graphical perception (1) review
  • Data typology
  • Expressiveness
  • Effectiveness
    • Signal detection
    • Magnitude estimation

Review

Sensory versus arbitrary

Preattentive attributes

Gestalt groupings

Leverage sensory and preattentive attributes to highlight specific messages and increase interpretability

Data typology

Data structure

  • Entities: objects of interest
  • Relationships: structures between entities
  • Entities and relationships both can be described with data attributes
  • Example

Data attributes

  • Nominal: no inherent ordering of values
    • Fruits: apples, bananas, oranges
  • Ordinal: ordered values with unknown distances between them
    • Quality: low, medium, high
  • Quantitative (interval): Only interval is meaningful
    • Date: 2/15/77, 3/18/66
  • Quantitative (ratio): zero is fixed
    • Physical measurements, counts, etc.

Expressiveness

A set of facts (data) is expressible in a language (layout) if:

  • It expresses all of the facts
  • It only expresses the facts in the set

Expressiveness is a function of:

  • The type of each data attribute of interest (N, O, Q)
  • The volume of data
  • The distribution of data

Unable to express all facts in a layout (fails first criterion)

Expresses information not inherent in the dataset (fails second criterion)

An alternative

Effectiveness

The effectiveness of a visualization is a measure of how accurately the values can be decoded

  • Prioritizes encodings that are most easily/accurately interpreted
  • Principle of Importance Ordering: Encode more important information more effectively (Mackinlay)

These rankings differ because of a variation in associated tasks

  • Nominal/Ordinal variables: detect differences
  • Quantitative variables: estimate magnitudes

Signal detection

Weber's law

  • Just noticeable differences are functions of relative (not absolute) difference
  • There is a constant scaling factor for interpretations of values (Weber's constant)

Magnitude estimation

Assessment

Error

Summary

  • Expressive graphics display all (and only) the facts of the data
  • Effective graphs exploit human perception to accurately communicate most important information
  • Leverage most perceptible features first (position,length, slope...)

Assignment 3 (conceptual)

  • Identify a global health dataset of interest
  • Sketch 5 representations of the dataset
  • Include short descriptions of how you encoded data attributes
  • Due Friday before class
  • Can be final project data
  • More details here