Parameters | Predictive Modeling | For Nominal Dependent Variables, compute distances to:

For Nominal Dependent Variables, compute distances to:
This feature enables you to specify a method for computing distances to either the centroid of each class or to every observation, and then consolidating results after the individual distances are computed.
The former can be considered a nearest centroid method, whereas the latter is weighted nearest neighbors with weights proportional to the inverse distances.
Note: This choice is available only when the dependent variable is nominal. If the dependent variable is continuous, then distances to each observation are always computed and summarized.
To Specify the Method for Computing Differences:
*