|
Activity Number:
|
24
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, July 31, 2016 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistics in Defense and National Security
|
|
Abstract #320376
|
View Presentation
|
|
Title:
|
On the Different Classes of Forensic Identification of Source Problems
|
|
Author(s):
|
Christopher Saunders* and Cedric Neumann and Danica Ommen
|
|
Companies:
|
MITRE/South Dakota State University and San Diego State University and South Dakota State University
|
|
Keywords:
|
Bayesian Inference ;
Model Selection ;
Forensic Science ;
Model Misspecification ;
Likelihood Ratios ;
Bayes Factors
|
|
Abstract:
|
In this presentation we will review the hierarchy of forensic propositions related to various source identification problems we have encountered while supporting the forensic science community. The purpose of this review will be to clarify the implied sampling experiments needed to discuss a formal likelihood structure for addressing the propositions of interest to the decision maker. Typically, an automated system is used to help provide an approximate likelihood ratio (LR) that a forensic scientist can provide as a quantification of the observed forensic evidence. These system-based LRs tend to be constructed to address a given proposition related to the testing and evaluation of the system in closed set identification, or the identification of a common but unknown source (to two sets of traces from an unknown source). Neither of these propositions tend to be of interest in a given court proceeding where there is a suspect at hand and all evidence interpretation is relative to the suspect as the known source. We will illustrate the various issues that arise due to this mismatching of propositions with examples from questioned document and trace element analysis.
|
Authors who are presenting talks have a * after their name.
