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

Area of Science:

• Psychology
• Criminology
• Sociology

Background:

• Investigating the link between psychological and physical aggression is complex.
• Previous research has utilized various methodologies to understand this relationship.

Purpose of the Study:

• To identify specific thresholds of psychological aggression at baseline.
• To differentiate individuals who will and will not perpetrate physical aggression at follow-up.

Main Methods:

• Employed unbiased recursive partitioning (conditional inference trees) on a longitudinal sample.
• Utilized anonymous self-report measures of psychological and physical aggression (CTS2) at baseline and 12-month follow-up.
• Categorized participants into risk groups based on aggression levels.

Main Results:

• The algorithm successfully categorized men into low- and high-risk groups for physical aggression.
• Women were categorized into mild-, moderate-, and high-risk groups.
• Sensitivity analyses demonstrated predictive accuracy for physical aggression up to 59% for women and 60% for men.

Conclusions:

• Conditional inference trees provide a robust method for identifying psychological aggression cutoffs predictive of physical aggression.
• Baseline psychological aggression is a significant predictor of future physical aggression perpetration in both men and women.
• This approach offers valuable insights for targeted interventions and risk assessment in relationship violence.