Online Program Home
My Program

Abstract Details

Activity Number: 342
Type: Topic Contributed
Date/Time: Tuesday, August 2, 2016 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics in Defense and National Security
Abstract #320296 View Presentation
Title: A Kernel-Based Method for the Forensic Inference of the Source of Small Particles Characterized Using Compositional Data
Author(s): Douglas Armstrong* and Cedric Neumann and Christopher Saunders and David Stoney
Companies: San Diego State University and San Diego State University and MITRE/South Dakota State University and Stoney Forensic
Keywords: Kernel Methods ; High Dimension ; Forensic Evidence ; Very Small Particles
Abstract:

The inference of source and quantification of the probative value of forensic evidence is a function of the type of evidence and of the analytical technique used to characterize it. In particular, many materials can only be characterized by using multiple, often expensive, analytical means. Assigning probability distributions to this heterogeneous data proves to be very complex.

Instead of looking at the evidence itself we look at very small particles (VSP) that attach to any evidence material. VSPs are gathered in the environment the trace material has been in and hold information on that environment. Thus, they can help infer the source of the evidence material. Importantly, the analytical technique characterizing the VSP is independent of the evidence type, reducing costs and increasing generalizability of the approach.

We use VSPs recovered from carpet fibers and apply current developments in kernel-based methods to compositional data to quantify the evidence and infer its source. This method can be easily extended to any type of VSPs found on other types of forensic materials such as weapons, drug packaging, or IEDs.


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

Back to the full JSM 2016 program

 
 
Copyright © American Statistical Association