Data mining in psychological treatment research: a primer on classification and regression trees

Area of Science:

• Psychology
• Statistics
• Computer Science

Background:

• Exploratory data analysis in treatment studies is underutilized.
• Traditional regression methods impose a priori relationships, limiting data-driven insights.
• Statistical learning offers an inductive approach, letting data suggest variable relationships.

Purpose of the Study:

• Introduce statistical learning, specifically classification and regression trees, as valuable tools for psychological research.
• Demonstrate how tree models can identify treatment effect predictors and interactions.
• Highlight trees' utility in clinical decision-making and hypothesis generation.

Main Methods:

• Classification and regression trees as exemplars of statistical learning.
• Inductive model building, contrasting with traditional regression.
• Application to randomized clinical trial data.

Main Results:

• Trees efficiently identify key predictors of treatment outcomes.
• Discovery of interactions between predictors without prior specification.
• Development of predictive models for clinical decision support.

Conclusions:

• Tree models offer a flexible and powerful alternative to traditional regression for analyzing treatment data.
• They excel at revealing complex patterns and informing hypotheses about treatment effects.
• Further exploration of tree models is recommended for treatment researchers.