Previous Page  Next Page 
Multivariate Analyses

Canonical Correlation Plots

You can request pairwise canonical variable plots and a plot of the first two canonical variables or the first three canonical variables from each variable set from the Canonical Correlation Options dialog, shown in Figure 40.8, or from the Graphs menu, shown in Figure 40.40.

[menu]
Figure 40.40: Canonical Correlations Menu

Figure 40.41 shows scatter plots of the first two pairs of canonical variables. The first scatter plot shows a high canonical correlation (0.7956) between canonical variables CX1 and CY1 and the second scatter plot shows a low correlation (0.2005) between CX2 and CY2.

mult36.gif (9412 bytes)

Figure 40.41: Canonical Correlation Pairwise Plots

Select Canonical Plot from the Canonical Correlations menu in Figure 40.40 to display a Canonical Correlation Component Plots dialog.

mult36a.gif (2973 bytes)

Figure 40.42: Canonical Correlation Component Plots Dialog

In the dialog, you choose a canonical correlation component scatter plot (Scatter Plot), a component biplot with standardized Y and X variables (Biplot (Std Y X)), or a component biplot with centered Y and X variables (Biplot (Raw Y X)).

In a canonical correlation component biplot, the data points are displayed in a scatter plot of canonical correlation components. With the approximated Y and X variable axes also displayed in the scatter plot, the data values of the Y and X variables are graphically estimated, as described previously in the "Principal Component Plots" section.

Figure 40.43 shows a biplot of the first two canonical variables from the Y variable sets with standardized Y and X variables. The biplot shows that the variables WEIGHT and WAIST (highlighted axes) have positive correlations with CY1 and negative correlations with CY2. The other four variables have negative correlations with CY1 and positive correlations with CY2.

mult37.gif (10934 bytes)

Figure 40.43: Canonical Correlation Component Biplot

Previous Page  Next Page  Top of Page

Copyright © 2007 by SAS Institute Inc., Cary, NC, USA. All rights reserved.