Small decisions about nothing can have a large effect on DNA transfer modelling for activity level evaluation

Area of Science:

• Forensic Science
• Statistical Modeling
• Biotechnology

Background:

• Activity level evaluations in forensic science depend on expert choices in simplifying complex real-world scenarios.
• These choices involve case circumstances, data selection, and data modeling techniques.
• Sensitivity analysis reveals that evaluations can be significantly influenced by any of these modeling decisions.

Purpose of the Study:

• To demonstrate the sensitivity of DNA transfer evaluations to the specific modeling approach chosen for DNA observations.
• To illustrate how different interpretations of 'no-DNA' observations can alter the strength of evidence supporting competing propositions.
• To propose a novel modeling solution that addresses the ambiguity in interpreting low-level or absent DNA signals.

Main Methods:

• Development of an evaluation scenario specifically designed to highlight the impact of DNA transfer modeling choices.
• Comparison of evaluation outcomes based on two distinct interpretations of non-DNA observations: complete absence versus below-limit detection.
• Implementation of a hierarchical Bayesian modeling approach to create a mixture of models.

Main Results:

• The choice of modeling DNA transfer, specifically how to interpret non-DNA observations, can shift an evaluation from slightly supporting one proposition to strongly supporting another.
• The proposed mixture model approach successfully integrates both interpretations (no DNA transfer and below detection limit) simultaneously.
• This integrated approach provides a more nuanced and robust evaluation of DNA transfer evidence.

Conclusions:

• The interpretation of non-DNA observations in DNA transfer experiments is a critical factor influencing evaluation outcomes.
• A mixture of models, facilitated by hierarchical Bayesian methods, offers a robust solution to account for the uncertainty in DNA transfer modeling.
• This approach enhances the accuracy and reliability of forensic evaluations involving DNA evidence.