Scatterplots of all pairs of variables
The problem of displaying relationships becomes even more difficult when there are more than three variables. It is possible to gain some insight into the relationships between the variables with scatterplots of all pairs of variables. If these scatterplots are arranged into an array, they are called a scatterplot matrix.
Brushing
Although a static display of a scatterplot matrix reveals some aspects of the relationships between the variables, more insight into the data is obtained by adding dynamic features.
All scatterplots are usually dynamically linked, so that clicking on a cross on one scatterplot highlights that individual in all scatterplots. This is often extended to allow highlighting of multiple crosses on a scatterplot with a 'brush' tool. Again, the crosses corresponding to these individuals are highlighted on all scatterplots. This is called brushing.
Dynamic linking of scatterplots and brushing are more easily explained with an example.
Swiss bank notes
The scatterplot matrix below shows four measurements that were made from 200 old Swiss 1000-franc bank notes. Two measurements describe the overall dimensions of the notes (the heights of the bills at their left and right edges) and two describe the printing within the bills (the heights of the margins at the top and bottom of the notes).
Various features are apparent ...
Do the clusters really represent different types of bank note? Are they distinct groups in the other two variables?
Drag with the mouse over the crosses on the bottom right scatterplot. Observe that the separate clusters on this plots have fairly similar height measurements (on the top left scatterplot).
Click the checkbox Use brush on the scatterplot matrix, then drag with the mouse over the crosses on the bottom right scatterplot. When all crosses in this cluster have been highlighted, look at the plot of the top and bottom margins (on the top left).
We will come back to this data set in a few pages.