Online Program Home
  My Program

Abstract Details

Activity Number: 423 - SPEED: Biopharmaceutical Statistics, Medical Devices, and Mental Health
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 2:00 PM to 2:45 PM
Sponsor: Social Statistics Section
Abstract #325313
Title: Assessing the Combination of Evidence Through a Probabilistic Graphical Model
Author(s): Amanda Luby* and Anjali Mazumder
Companies: Carnegie Mellon University and Carnegie Mellon University
Keywords: Forensic Science ; Graphical Model ; Evidence
Abstract:

In forensic science, analysts are often tasked with addressing what is known as a 'source' or 'sub-source' proposition using the evidence available. More recently, analysts have been asked to assess 'activity' level propositions, which go beyond identifying or confirming a suspect and begin to address the likelihood of observing certain evidence under different scenarios. This often involves combining multiple types of evidence. We have defined three classes of complex forensic problems which involve combining different types or items of evidence in a single case. These classes or scenarios include: (1) combining different aspects of the same evidence type to address a single proposition; (2) combining different types of evidence to address the same proposition level; and (3) combining evidence from the same item to address different proposition levels. Some of the issues we address are: independence (or lack thereof) between evidence types, the hierarchical nature of propositions, incorporating expert judgement, and formalizing uncertainty. We believe that the most effective method for doing so is to expand upon work using a probabilistic graphical model representation.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association