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.
Characteristics of supermarket shoppers
A supermarket conducts a survey of customers to help identify the characteristics that it should target in its advertising. The jittered dot plot on the right below shows the annual incomes of 100 female shoppers. The scatterplot on the left shows the weekly purchases of groceries and age of the respondents earning between $40,000 and $60,000 per year.
Drag the red band on the jittered dot plot up and down to display different slices of the individuals (rich, average or poor) 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.
Characteristics of supermarket shoppers
The trellis plot below shows the relationship between parchases and age for various income-slice.