Improving calibration of forensic glass comparisons by considering uncertainty in feature-based elemental data

Forensic glass examinations—A review focused on elemental spectrochemical analysis

“…A drawback of the comparison interval methods is that they do not assess the glass composition's rarity, and the results are prone to the fall‐off‐the‐cliff effect (Buzzini & Curran, 2020; Curran et al, 2020; Ramos et al, 2021). Alternatively, Bayesian approaches provide additional advantages to glass interpretation.…”

“…The low rates of misleading evidence assessed the LR approach's performance (RoME < 1.5%; Corzo et al, 2018). Ramos et al discussed in another manuscript the basis and performance of the developed algorithm (Ramos et al, 2021). In a follow‐up study, Gupta et al investigated the use of PCA for dimensionality reduction of multivariate elemental concentrations of glass for computing likelihood ratios and compared it to the multivariate kernel‐PVA calibration previously reported.…”

“…The calibration of LR systems is a means to assess the performance of LR systems and avoid underestimating or overestimating the strength of the evidence. In a more general context, researchers and practitioners continue discussing the need for calibrating likelihood ratios and evaluating the optimal approaches for calibration (Hannig & Iyer, 2022; Malmborg & Nordgaard, 2022; Ramos et al, 2013; Ramos et al, 2021; Vergeer et al, 2021; Vergeer, Alberink, et al, 2020). Malmborg et al described SAILR, an open‐source software designed to calculate forensic likelihood ratios from probability distributions of reference data.…”

Atomic spectrometry update: review of advances in the analysis of metals, chemicals and materials

“…A related paper, by Ramos et al , described the probabilistic two-level modelling of the within-source and between-source variability that is required for the calculation of likelihood ratios. 336 The probabilistic machine learning algorithms used comprised both a variational autoencoder and a warped Gaussian mixture. It was stressed that the model developed had significant advantages over previous models used to calculate likelihood ratios.…”

Interpol review of paint, tape, and glass evidence 2019–2022

“…Ramos et al [ 191 ] propose models that compare favorably to previously proposed feature-based LR models, by improving the calibration of the computed LRs based on quantitative elemental analysis using LA-ICP-MS and the ASTM E2927 method. The authors assume that the within-source variability is heavy-tailed to incorporate uncertainty when the available data is scarce as it typically happens in forensic glass comparisons.…”