ABSTRACT
Trace evidence analysis has been used in forensic science to analyze the source of physical trace evidence (bullets, glass, paint, etc.) under the working hypothesis that elemental concentrations may provide a “signature” allowing one to determine whether two pieces of evidence are “analytically indistinguishable.” Recently, three American Society for Testing and Materials (ASTM) International standards have been proposed or adopted for comparing glass fragments found at a crime scene to those recovered from a potential suspect. Although error rates from referenced papers are stated to be less than 1%, they are calculated from highly diverse glass databases that are not necessarily representative of glass fragments seen at a crime scene.
We describe statistical approaches for estimating error rates in the proposed procedures to assess analytical indistinguishability that do not depend on specific glass collections. Procedures proposed by ASTM should undergo more rigorous validation testing to estimate more realistic error rates independent of any population of glass. This is especially critical for trace evidence that may potentially be contaminated by environmental factors.
