Correlation Radius for Clustering

Use this option to specify a correlation value to be used to determine the minimum radius for clustering. This value must be between 0 and 1. The higher the correlation, the more clusters that are defined with members of each cluster being more highly correlated with one another than with other clusters. A value of 8 is specified by default.

Note: This criterion is disabled when a value of 0 is specified.

Specifying this correlation as rho (r), sets the radius for SAS PROC FASTCLUS equal to:

where MeanUSS is the average of row-wise uncorrected sums of squares.

To Specify a Correlation Radius for Clustering:

8

Check the Use K-Means clustering to reduce predictors check box.

8

Either highlight the existing value and enter the desired value in the text field or move the slider to the right or to the left to change the value.

Tip: To change the scale of the slider, right-click on the slider and select Rescale Slider from the pop-up menu. Change the upper and/or lower boundaries in the window that appears, to rescale the slider and click OK.