Customize Groups for a Categorical Variable

  1. Select a categorical variable in the Predictor variables pane in the Weight of Evidence (WoE) Startup Panel.
  2. Click the Customize groups button to display the Customize Groups for a Categorical Variable dialog, which contains options to select a custom grouping for the discrete values of the predictor.

    By default, the modified CHAID algorithm (described in the Algorithm section of the WoE Introductory Overview) is used to create a default best grouping for the selected variable.

Merging Groups

  1. To merge groups, click on the respective rows (groups) in the Categories pane.
  2. Click the Group button.

Results:

  • Merged groups
    • display with identical colors (red, or green)
    • show common WoE, Gini, and contribution to the Information Value (IV) statistics
  • The overall Gini and Information Value (IV) will update.

Splitting (UnGrouping) Groups

To remove one or more groups from a grouping:

  1. Highlight the respective groups by clicking on the respective rows in the Categories pane.
  2. Click the UnGroup button. All statistics will be updated.

Group

  1. Select rows (groups) in the Categories pane.
  2. Click the Group button to merge the selected groups.

UnGroup

To remove the selected group:

  • Select a group in the merged groups.
  • Click the UnGroup button.

Clear Selection

To clear the selected groups in the Categories pane, click the Clear Selection button.

OK

To save the current grouping, click the OK button.

Cancel

To abandon the current action, click the Cancel button

Comments

Add or edit comments in this box. Only use  text  and do not exceed 2,000 characters. The comments will be displayed in the header of the custom crosstabulation output.