Grammar of Graphics

Andrew Irwin, a.irwin@dal.ca

Math & Stats, Dalhousie University

2021-09-20

Development

Leland Wilkinson

Hadley Wickham

What is the grammar of graphics?

  • A way of describing a visualization

    • Data
    • Mapping from data to aesthetic features
    • Geometry
    • Scales
    • Facets
    • Statistics
    • Coordinates
    • Themes

What is the grammar of graphics?

Data

We will use “tidy data” arranged in a table

  • columns are variable
  • rows are observations

This arrangement makes it easy to use one variable for each aesthetic feature of a plot (axis position, colour, facet, text labels, …)

Aesthetic mappings

We map variables to features of the graph

  • x, y axis position
  • colour (outline, fill)
  • symbol shape, size
  • line type
  • error bars
  • many more used less frequently, depending on the geometry

Discrete and quantitative variables are handled the same way. The details of the scale and geometry will determine their interpretation.

Scales

  • Range of values on x or y axis; formatting of axis labels

  • Mapping from values to a discrete or continuous colour scale; choice of colours

  • Mapping from values to a shape, or line thickness, or symbol size

Geometry

What kind of plot do you want?

Facets

Statistics & Coordinates

  • Do you want to perform calculations on your data before plotting?

    • Calculate average and standard deviation
    • Smooth data
    • Add a regression line

  • What coordinate system do you want to use?

    • Cartesian (x-y)
    • Polar (r, \(\theta\))
    • Map projections

Annotation & Theme

All the “non-data ink”:

  • Titles, axis labels, tick marks
  • Background shading, box around plot
  • How and where the legend (colour, shape) is displayed
  • Other annotations - shading, text

Summary

  • The grammar of graphics is a way of describing a visualization.

  • ggplot2 is a computer implementation of this idea.

  • Other implementations are available for python (plotnine), javascript (D3), julia (Gadfly).

Suggested reading

  • Course notes

Self-quiz / Bonus task

Aesthetics are mappings between variables in your dataset and your visualization, defined using the aes function.

Which of the following is a possible aesthetic mapping for the gapminder data set?

  1. aes(country = x, pop = y)
  2. aes(x = country, y = pop)
  3. aes(x <- country, y <- pop)

Self-quiz / Bonus task

The geometry (or geom) of a plot determines how the data are displayed in your visualization.

Select the true statement.

  1. You can have only one geometry, such as geom_point per plot
  2. You can use mutiple geoms per plot, for example combining both geom_plot and geom_line.

Self-quiz / Bonus task

True or false: To change all the points in a scatterplot to red circles, you need to add variables to your data frame with values “red” and “circle” and map those variables to the colour and shape aesthetic.