Two‐stage approach for the inference of the source of high‐dimensional and complex chemical data in forensic science

Abstract

Forensic chemists are often criticised for the lack of quantitative support for the conclusions of their examinations. While scholars advocate for the use of a Bayes factor to quantify the weight of forensic evidence, it is often impossible to assign the necessary probability measures to perform likelihood‐based inference on chemical data. To address this issue, we leverage the properties of kernel functions to offer a method that allows for statistically supporting the inference of the identity of source of sets of trace and control objects by way of a single test. Our method is generic in that it can be easily tailored to any type of data encountered in forensic chemistry, and our method does not depend on the dimension or the type of the considered data. The application of our method to paint evidence analysed by FTIR shows that this type of evidence carries substantial probative value. Finally, our approach can easily be extended to other types of chemical evidence such as glass, fibres, and dust.

Defence against the modern arts: the curse of statistics—Part II: ‘Score-based likelihood ratios’

“…We are not arguing that these ad-hoc methods are not useful in their own way, but the harsh reality is that if one wants to abide by the idea that forensic evidence should be reported within a Bayesian paradigm, then one cannot use score-based likelihood ratios. We appreciate that the use of scores may be the only viable method to reduce the complexity of forensic evidence, but more efforts should be spent in the development of more rigorous models for handling these scores (see (Armstrong et al, 2017;Ausdemore et al, 2019;Hendricks et al, 2019) for some early work on these models).…”

“…We warn the reader that other models involving scores have been proposed (Armstrong et al, 2017;Swofford et al, 2018;Ausdemore et al, 2019;Hendricks et al, 2019) but are not considered to be score-based likelihood ratios. These models do not rely on the ratio of the likelihoods of the score in two sampling distributions.…”

Use of an Automated System to Evaluate Feature Dissimilarities in Handwriting Under a Two‐Stage Evaluative Process*†

“…When failing to conclude that the two suspect samples do not share a common source, the two-stage approach requires the use of evidence from a specified background population to assess the evidential support for the proposition that the suspect samples do share a common source. For an overview on the comparison of likelihoodbased and two-stage approaches, see (22); for a discussion as it applies to handwriting examination, see (21); and for a rigorous statistical discussion on two-stage approaches for Bayesian model selection as it pertains to specific-source propositions, see (23).…”

Innovative Vibrational Spectroscopy Research for Forensic Application

“…This method uses kernel functions in two steps to assign a probative value to the results of paint evidence: (1) by testing for indistinguishability and (2) by assessing the probability of a match. 18 Other work has focused on further ways to analyze collected spectral information. de Abreu Fontes et al proposed a method for analyzing near-IR (NIR) spectra.…”