This article explains the output of "predictor stats" generated from Interactive Trees Module.

For illustration purpose, the example data CreditScoring.sta is applied by the analysis of "Interactive Trees | Regression Analysis | Exhaustive CHAID".

1. After the tree is grown, click "Predictor stats" under the Itree CHAID Results dialog. To check predictor statistics for other nodes, change the Node id.

Note: click "?" on top right to access e-manual explaining every option in this dialog. In the e-manual , find the "Predictor stats" section and follow the link for more details about the computation of the statistics.

2. A spreadsheet with the predictor statistics for the current node(Node 1 in this example) will be generated. The predictors are ordered according to Statistic/df (shown in the spreadsheet title), not necessarily lined up with the rank of the significance of the splitting predictor. The predictor with smallest adjusted p-value is selected as the best splitting predictor for current node.

3. To check the exact value of adjusted P-value, double click the column of the adjusted P and change its display format to be scientific notation.

Or to rank the significance of predictors, you can sort the spreadsheet by the adjusted P under "Data | Sort".