Multivariate Analysis: Discriminant Analysis

The Plots Tab

You can use the Plots tab to create plots that graphically display results of the analysis. (See Figure 30.4.)

Creating a plot often adds one or more variables to the data table. The following plots are available:

Observed groups

creates a spine plot (a one-dimensional mosaic plot) of the groups for the Y variable.

Observed vs. Predicted groups

creates a mosaic plot of the groups for the Y variable versus the group as classified by a discriminant function. Each observation is placed in the group that minimizes the generalized squared distance between the observation and the group mean.

Classification fit plot

creates a plot that indicates how well each observation is classified by the discriminant function. This plot is shown in Figure 30.5. The observations selected in the plot have a low posterior probability of group membership.

For each observation, PROC DISCRIM computes posterior probabilities for membership in each group. Let $m_ i$ be the maximum posterior probability for the ith observation. The classification fit plot is a plot of $-\log ( m_ i )$ versus i.