Developing a Short-Form Buss-Warren Aggression Questionnaire Based on Machine Learning.

Area of Science:

• Psychology
• Machine Learning
• Adolescent Health

Background:

• High adolescent aggression correlates with negative outcomes like crime and suicide.
• The Buss-Warren Aggression Questionnaire (BWAQ) is a lengthy 34-item tool.
• Longer questionnaires risk insufficient effort responses (IER) and increase costs.

Purpose of the Study:

• To develop a shorter, more efficient aggression questionnaire using machine learning (ML).
• To reduce insufficient effort responses (IER) and implementation costs.
• To maintain measurement accuracy and validity in a condensed format.

Main Methods:

• Developed an initial short-form questionnaire using stepwise regression and ANOVA F-test.
• Utilized a machine learning algorithm to create the optimal short-form questionnaire (BWAQ-ML).
• Validated the BWAQ-ML's reliability and performance using independent samples.

Main Results:

• The BWAQ-ML comprises only four items, significantly reducing length.
• Achieved high performance metrics: AUC (0.85), accuracy (0.85), recall (0.89), precision (0.83), and F1 score (0.86).
• Demonstrated good reliability (Cronbach's alpha = 0.84) and construct validity.

Conclusions:

• The BWAQ-ML is an effective and efficient tool for measuring adolescent aggression.
• Its brevity improves efficiency for large-scale assessments and minimizes IER.
• Facilitates early identification and intervention for aggressive adolescents.