|
Abstract:
|
In the US forensic science community, NIST is leading an effort to regularize the statistical methods used for reporting and interpreting scientific evidence. The results of this effort will also impact the use of these and similar methods for detecting potential threats in the intelligence community. The leading method for quantifying the probative value of evidence for identification of source is the subjective Bayesian approach of providing a Bayes Factor, which is a statistic that summarizes the information contained in the observed data under two competing hypotheses. Once the value of evidence has been quantified, it can be used to update a decision-maker's belief about the relative merit of the competing hypotheses and ultimately make a decision as to which hypothesis to support. Currently, the computationally intensive nature of the Bayesian approaches limits their applicability in practice. Thus, research into approximate solutions for identification of source problems has become a major focus in forensic statistics. In this presentation, we will present a number of computationally efficient information criteria based methods for quantifying the value of evidence.
|
