IBM Support

Feature Selection: Computation of Importance Value using Cramer's V or Lambda uses the value of the measure of association instead of the p-value

Troubleshooting


Problem

I am using the Feature Selection node in Clementine to identify fields that are most important in my dataset for a given analysis. All predictors and the target within my dataset are categorical variables, therefore I want importance to be ranked based on Cramer's V measures of associate. Clementine helps indicates that Importance is measured on a percentage scale and can be defined broadly as 1 minus the p value of the measure of association chosen. However, a look at the importance value returned by Clementine indicates that the computation of the Importance Value is based on the statistic of the measure of association (i.e. Cramer's V) instead of the p-value the statistic. Why is this?

[{"Product":{"code":"SS3RA7","label":"IBM SPSS Modeler"},"Business Unit":{"code":"BU048","label":"IBM Software"},"Component":"Modeler","Platform":[{"code":"PF025","label":"Platform Independent"}],"Version":"13.0","Edition":"","Line of Business":{"code":"LOB76","label":"Data Platform"}}]

Log InLog in to view more of this document

This document has the abstract of a technical article that is available to authorized users once you have logged on. Please use Log in button above to access the full document. After log in, if you do not have the right authorization for this document, there will be instructions on what to do next.

Historical Number

71057

Document Information

Modified date:
16 June 2018

UID

swg21478681

Page Feedback