Optimal Binning Startup Panel - Advanced Tab

Select the Advanced tab of the Optimal Binning Startup Panel to access the options described here. As described in the Introductory Overview, the program uses a CHAID-like algorithm to find a combination of predictor categories that strongly relates to the dependent variable of interest. For details regarding this algorithm, refer to  the documentation for General CHAID Models (GCHAID) and Interactive Trees (C&RT, CHAID). The options described here are identical to those discussed there, e.g., on the General CHAID Models Quick Specs dialog box - Stopping tab.

Element Name Description
Stopping parameters The parameters specified in this group box determine when the codes-aggregation process will terminate.
Min-N to stop (% of cases) Enter a percentage value in this field to specify when the split selection (combining groups into aggregated or recoded new groups) should stop. The percentage value entered here will be evaluated against the total number of observations (cases) in the input spreadsheet, regardless of missing data. For example, if the input file contains 100,000 cases and you entered 5% in the Min-N to stop (& of cases) field, the minimum N for a recoded group would be set to .05*100,000 = 5,000.  If the number of observations within a node is less than this value, the node will not be considered for further splitting.
Prob. for splitting. As described in Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID, the split selection (process of combining predictor classes) will continue until no further splits can be found that are statistically significant (see also, Bonferroni adjustment, below) at the level specified here as the Probability for splitting.
Prob. for merging. As described in Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID, the computations involve the merging of predictor categories that are not statistically significant with respect to the dependent variable (chi-square tests are computed for categorical dependent variables, and F tests are computed for continuous dependent variables). If no further merging can be performed, giving the Probability for merging defined here, and if no further splits can be performed consistent with the other stopping criteria specified on this tab, then the split selection computations (process of combining predictor classes) will terminate.
Splitting merged categories When a predictor is used in the splitting criterion, the CHAID analysis merges the current categories into as small a number of categories as possible in order to find a parsimonious split rule for the final aggregation and recoding of predictor codes. If this check box is selected, the merged categories are split to optimally select the categories. See also Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID for details.
Bonferroni adjustment As described in Basic Tree-Building Algorithm: CHAID and Exhaustive CHAID, at the point of selecting the best predictor for a split, the program finds the predictor with the smallest p-value (greatest statistical significance) for the set of categories for the respective predictor. This p-value can be computed after applying the Bonferroni adjustment.