This is the final summary overview report on the findings and methodology of a project that examined the validity, accuracy, and computational complexity of methods designed to quantify the weight of complex evidence forms, such as pattern evidence and trace evidence.
The overall conclusion of this report is that the project has "extended the ability of forensic statisticians to quantify the value of complex evidence forms in the rigorous, logical, and coherent manner advocated by legal and scientific scholars for the past four decades." Five criminal justice benefits of this project are outlined. First, it partitioned the identification of source in forensic science using two distinct formal frameworks. Second, it showed that a popular method for "quantifying" the weight of forensic evidence, namely "score-based likelihood ratios," is not appropriate and could result in misleading evidence. Third, it proposes different methods for addressing issues related to the calculation of formal Bayes factors, namely, the difficulty in assigning posterior probabilities to models' parameters and the error associated with Monte Carlo integration. Fourth, since the aforementioned methods only apply to situations in which the evidence can be characterized in a low dimensional feature space, the current project proposes two methods that leverage the power and flexibility of kernel function, so as to facilitate the probabilistic interpretation of forensic evidence. Fifth, the project trained six PhD and MSc graduate students in the interpretation of forensic evidence by using probabilistic models. The project relied on a mixture of analytical proofs and statistical simulations in developing and studying numerical methods for assigning the weight of forensic evidence. Project phases are described in detail. 3 figures and 12 references
- Social Preferences as an Individual Difference in Offender Decision-making
- No Escape: Mass Incarceration and the Social Ecology of Intimate Partner Violence Against Women
- Distinguishing Infested Flour from Uninfested Flour through Chemometric Processing of DART-HRMS Data─Revealing the Presence of Tribolium castaneum, the Red Flour Beetle