Usage of K-Means Clusters

Use this option to choose how to use the clusters.

Options for how to use the clusters to reduce the number of predictors are described in the table below:

Option	Description
Select One representative predictor from each cluster	Choose this option to remove predictor redundancy. The representative is chosen as the predictor with closest Euclidean distance to the cluster mean.
Use the cluster means themselves as predictors	Choose this option to replace all of the original predictors. This approach can be more difficult to interpret since it is a function of the original predictors, but it avoids potential bias from any one predictor.

Option

Description

Select One representative predictor from each cluster

Choose this option to remove predictor redundancy.

The representative is chosen as the predictor with closest Euclidean distance to the cluster mean.

Use the cluster means themselves as predictors

Choose this option to replace all of the original predictors.

This approach can be more difficult to interpret since it is a function of the original predictors, but it avoids potential bias from any one predictor.

To Specify How the K-Means Clusters Are to Be Used:

Check the Use K-Means Clustering to Reduce Predictors box.

Select the desired method by clicking within the appropriate radio button.