Slicing a scatterplot using a third variable

Brushing crosses on a scatterplot highlights selected individuals on all linked displays. Slicing is a closely related dynamic technique, the difference being that individuals that are not 'selected' are not displayed at all.

The individuals that are selected are usually those whose values for one variable are within a specified range. The selection therefore defines a slice of the distribution of that variable. The selection range can be adjusted to display different slices of the variable's distribution.

The idea is easy to understand with an example.

Body fat

The jittered dot plot on the right below shows the heights of 252 men. The scatterplot on the left shows the body fat and weight for the individuals whose heights are in the band between 70 and 74 inches.

Drag the red band on the jittered dot plot up and down to display different slices of the individuals (short, medium or tall) in the scatterplot.

You should observe that ...

Trellis plot

Slicing is a dynamic method and must therefore be performed on a computer. Although less effective, an alternative is to show a series of static plots for different slices of the data. This is called a trellis plot.

Body fat

The trellis plot below shows the relationship between body fat and weight for various height-slice.