Posterior distributions for likelihood ratios in forensic science.

Area of Science:

• Forensic Science
• Statistical Inference

Background:

• Traditional forensic evidence evaluation often integrates or conditions on parameters, potentially eliminating uncertainty.
• The posterior likelihood approach retains uncertainty by considering parameter uncertainty derived from posterior distributions.

Purpose of the Study:

• To evaluate forensic evidence by retaining uncertainty in likelihood ratios using posterior distributions.
• To address criticisms regarding the applicability of the posterior likelihood approach and illustrate its practical use.

Main Methods:

• Utilizing posterior distributions to represent uncertainty in likelihood ratios.
• Summarizing posterior distributions for likelihood ratios using the median and credible intervals.
• Applying the method to forensic data for body height estimation.

Main Results:

• Posterior distributions effectively retain uncertainty in likelihood ratios.
• The median and credible intervals are recommended for summarizing these distributions.
• The posterior mean is not recommended for summarizing the distribution.

Conclusions:

• The posterior likelihood approach offers a valuable method for forensic evidence evaluation by preserving uncertainty.
• The study demonstrates a practical application in body height estimation, addressing concerns about applicability.
• Further exploration of this method can enhance the rigor of forensic statistical analysis.