Skip to main content

Multivariate Data Visualisation

Intuition

  • 2 geometric (position) display dimensions
  • For data sets with >2 variables, we must project data down to 2D
  • Come up with visual mapping that locates each dimension into 2D plane

Two ways to present the data:

  1. Directly (Textually): Tables
    • Limitations: Occupy large space, difficult to understand the relationships. Hard to see the overall picture, focus and see the context
    • When to Use: Look up individuals values, compare , precise values
  2. Symbolically (Pictures): Graphs
    • When to use: message contained in the shape of the values. Reveal relationships. Number of data points is huge

Multivariate Data Visualisation

Strategies:

  • Avoid "over-encoding"
  • Use space and small multiples intelligently
  • Reduce the problem space
  • Use interactions to generate relevant views

Rarely does a single visualisation answer all questions. Instead, the ability to generate appropriate visualisations quickly is key

Chernoff Faces

  • Encodes different variables values in characteristics of human faces

Tables Lens

  • Spreadsheet is certainly one hypervariate data presentation
  • Make the test more visual and symbolic

Parallel Coordinate

  • Encode variables along horizontal row
  • Vertical line specifies different values that variable can take
  • Data point represented as a polyline

Mosaic Plot

  • Categorise multiple things at once

Visualisation Tools

Grammar of Graphics

  • Programming by describing what, not how
  • Separate specification from execution
  • Faster Iteration Better visualisation Reuse Performance Portability Programmatic generation