An Exploratory Technique for Investigating Large Quantities of Categorical Data

Abstract¶

The technique set out in the paper, chaid, is an offshoot of aid (Automatic Interaction Detection) designed for a categorized dependent variable. Some important modifications which are relevant to standard aid include: built-in significance testing with the consequence of using the most significant predictor (rather than the most explanatory), multi-way splits (in contrast to binary) and a new type of predictor which is especially useful in handling missing information.

Summary¶

Introduces CHAID (Chi-square Automatic Interaction Detection), a multi-way decision tree algorithm for categorical outcomes that uses chi-square tests to select split variables and allows more than two branches per node. CHAID pioneered statistical significance testing as the split criterion, addressing AID's bias toward high-cardinality variables.

Links¶

Tags¶

Similar Papers¶