Scatterplot matrices are an excellent architecture for displaying large amounts of hypervariate data. Edward Tufte named this type of "multi-window plot" the most ingenious design. William Cleveland proposed giving an award to the (unknown) inventor of the scatterplot matrix and here is what he had to say about this graphical method:
"The inventor may not have fully appreciated the significance of the method or may have thought the idea too trivial to bring forward, but its simple, elegant solution to a difficult problem is one of the best graphical ideas around for displaying scattered measurements of three or more variables."
The basic concept of the scatterplot matrix is simple. Given a set of N variables, the scatterplot matrix is a N*N matrix containing all the pairwise scatterplots of the N variables in a single eye span—hence enabling comparisons.
Unfortunately, Excel charting library doesn't include native small multiples capabilities. So let’s build one in Excel. We're looking here at the famous Auto MPG data set. The data consist of three variables, that is, mpg, horsepower and acceleration measured for 398 cars. The matrix is made of 3*3 panels. Each panel of the matrix (except the diagonals) is a scatterplot of one variable against another.
The first step of the process is to create a scatterplot and add to it the vertical and horizontal grids that define the 9 panels. Since we have 3 panels in each direction we need to create four series—with 2 data points each—and crossing the axes at 1/3 (or 0.3333) and 2/3 (or 0.6666) as per the following setup: