Calculating Probability in Sex Offender Risk Assessment

Area of Science:

• Forensic Psychology
• Biostatistics
• Behavioral Sciences

Background:

• Probability quantification is essential in biomedical and behavioral sciences.
• Bayesian and Frequentist models define probability differently.
• Bayesian and regression models are used for probability quantification, yielding varied results.

Purpose of the Study:

• To compare Bayesian and regression models for quantifying probability.
• To evaluate the utility of the Bayesian model for assessing individual recidivism risk.
• To propose a forensic practice guideline for risk assessment.

Main Methods:

• Comparison of Bayesian probability quantification with logistic regression.
• Demonstration of calculating Bayesian probabilities and credible intervals from actuarial data.
• Development of a guideline for interpreting risk thresholds.

Main Results:

• The Bayesian model is presented as a viable alternative to logistic regression.
• The Bayesian model may be more effective for quantifying absolute recidivism risk in individual sex offenders.
• Credible intervals can be easily calculated from actuarial data sets.

Conclusions:

• The choice between Bayesian and regression models for probability quantification remains controversial.
• The Bayesian model offers a practical and potentially more useful approach for individual risk assessment.
• A guideline is proposed to ensure risk assessments are based on credible margins of error.