Campus Units

Statistics, Center for Statistics and Applications in Forensic Evidence (CSAFE)"

Document Type

Poster

Conference

American Association of Forensic Sciences (AAFS)

Publication Date

2021

Abstract

Background: LR are advocated as a way to provide numeric assessment of the evidential strength during expert's testimony. Researchers can implement SLR as way to reduce a complex model into a lower dimensional metric and avoid distributional assumptions1. Objective: Using simulated data to i) compare the discriminative power of SLR. ii) Compare their behavior w.r.t. to LR. iii) Assess the dependence on training data. Data: Simulated datasets2 based on glass data3. In each set: 10 sources with 5 fragment each and 3 features. Pairwise comparisons between fragments create a database of 100 KM and 1125 KNM.

Comments

Posted with permission of CSAFE.

Copyright Owner

The Authors

Language

en

File Format

application/pdf

Share

 
COinS