Conditional inference trees: a method for predicting intimate partner violence.

A number of different methodologies have been employed to investigate the complex relationship between psychological and physical aggression. Herein, a method of unbiased recursive partitioning (conditional inference trees) was applied to a longitudinal sample to identify cutoffs of psychological aggression at baseline that differentiate between individuals who do and do not perpetrate physical aggression at follow-up. The algorithm categorized men into low- and high-risk groups, and women into mild-, moderate-, or high-risk categories of perpetration. Couples responded anonymously to a self-report measure of psychological and physical aggression (CTS2) at baseline and a 12-month follow-up. Sensitivity analyses for predicting physical aggression reached as high as 59% for women and 60% for men.