📸 Decision Tree (Classify → Tree) menu in SPSS
What Is a Decision Tree?
A decision tree partitions data into subgroups using a series of binary splits based on predictor variables, building a tree-shaped classification or prediction model. Each internal node represents a split rule; each leaf node represents an outcome class. Decision trees are highly interpretable — even non-statisticians can follow the logic.
CHAID vs. CRT
- CHAID: Chi-square based, can create multi-way splits, good for categorical outcomes.
- CRT (Classification and Regression Trees): Binary splits only, handles both categorical and continuous outcomes, based on Gini impurity.
Running in SPSS
Step 1: Analyze → Classify → Tree.
Step 2: Add DV to Dependent Variable. Add all candidate predictors to Independent Variables.
Step 3: Growing Method: CHAID or CRT. Set Maximum tree depth (5 is a good starting point).
Step 4: Validation: Cross-validation (10-fold) or holdout sample (70/30 split) → OK.
📸 Classification matrix — 85.5% overall accuracy
APA Reporting
CHAID decision tree analysis achieved 85.5% overall classification accuracy, with 88.8% specificity and 80.9% sensitivity in cross-validation. The root node predictor was CRP level (χ²=42.3, p<.001), explaining the largest proportion of outcome variance.