Skip to content

Set the CHAID options

The CHAID algorithm creates splits by grouping together categories based on the P value of Chi Squared tests. These options are set on the Advanced Options tab of the Build Options.

Reducing the "Minimum p value to Merge" value will tend to lead to fewer branches.

Overview of algorithm

  1. Initially all categories are considered to be in separate branches.

  2. Effectively, a Chi Square test is done for every combination of pairs of branches. This is a 2 x 2 test of the two branches versus the binary condition defined by the analysis selection.

  3. The pair of branches which are most similar are joined together to form a new branch, if the P value from this test is greater than the "Minimum p value to Merge".

  4. Stages 2 and 3 are repeated until no more branches merge.

  5. A Chi Square test is done for the set of branches versus the binary condition defined by the analysis selection.

  6. The dimensions are ranked according to the P value from this Chi Square test. In practice the Significance figure is used, which is a log measure, since P values are typically very small.

  7. If the top ranking split passes the Split Validation conditions, then it is used to split the node. The "p value to split" measure used in Chaid is essentially the Significance P value in these conditions.

  8. The next node to split is by default the node which offers the "Best Split". This is the node whose top ranking split has the lowest P value (highest Significance value).