Data density

Estimate the number of data entries per square inch of a figure. This implements Tufte's data-density measure. It can help you compare how much information different versions of a figure occupy on the page.

Usage

data_density(plot, width = 6.5, height = 4, panel_only = TRUE)

Arguments

plot

A ggplot object.

width, height

Intended printed size in inches. Defaults to 6.5 by 4.

panel_only

Logical. Measure against the panel area rather than the whole figure? Defaults to TRUE, which is what Tufte means by "the data graphic". The panel share is estimated by rendering the plot.

Value

An object of class tufte_density: a list with density (entries per square inch), entries, rows, variables and area.

Details

Entries are counted as pooled rows times the number of distinct variables mapped to aesthetics. Constants outside aes() don't count. Statistical layers and plots that combine data sources need care: the function doesn't reconstruct a separate data matrix for each layer.

Panel area is estimated from the rendered plot. If that estimate isn't available, the function uses the whole canvas. I'd read the result alongside the entry count and the figure, since a high density doesn't establish that the information is useful.

Examples

library(ggplot2)
data_density(ggplot(mtcars, aes(wt, mpg)) + geom_point())
#> 
#> ── Data density 
#> 3.1 numbers per square inch of data graphic.
#> • 32 rows x 2 mapped variables = 64 entries
#> • over 20.83 square inches