arXiv subjects

Amy L Wilson

Publications and source records attributed to Amy L Wilson.

3 recordsLinked to original sources

arXiv2026

Evaluating the probative value of forensic gait analysis evidence using empirical data

Forensic gait analysis can aid the investigation of crimes through comparing features of gait captured in video footage. Modelling the probative value of gait evidence requires an understanding of the variation of features of gait between individuals in the population and within the same individuals. We address this question using a previously described population dataset and newly collected datasets with repeated observations of the same individuals on separate occasions. In addition to exploring the level of variability, correlation between features of gait, and the effect of demographic factors, we developed a likelihood ratio model through recoding features of gait as dichotomous variables and dimension reduction using PCA. High correlations between some features were observed, confirming that they should not contribute independently to the weight of evidence. The likelihood ratio model produced misleading likelihood ratios in less than 10% of the comparisons using the first four principal components. However, the risk increases when within-individual variability is mis-specified. Therefore, while the current model provides assistance to the judgement of gait experts, human expertise is indispensable to decide whether or not the difference in walking and/or recording conditions between the reference and questioned footage could have caused any observed differences in the features of gait. We discuss future directions in understanding the sources of the variability, improving statistical modelling and note the need to consider carefully how to select the relevant population for model fitting.

stat.AP↗

arXiv2024

Chain event graphs for assessing activity-level propositions in forensic science in relation to drug traces on banknotes

Graphical models and likelihood ratios can be used by forensic scientists to compare support given by evidence to propositions put forward by competing parties during court proceedings. Such models can also be used to evaluate support for activity-level propositions, i.e. propositions that refer to the nature of activities associated with evidence and how this evidence came to be at a crime scene. Graphical methods can be used to show explicitly different scenarios that might explain the evidence in a case and to distinguish between evidence requiring evaluation by a jury and quantifiable evidence from the crime scene. Such visual representations can be helpful for forensic practitioners, the police and lawyers who may need to assess the value that different pieces of evidence make to their arguments in a case. In this paper we demonstrate for the first time how chain event graphs can be applied to a criminal case involving drug trafficking. We show how different types of evidence (i.e. expert judgement and data collected from a crime scene) can be combined using a chain event graph and show how the hierarchical model deriving from the graph can be used to evaluate the degree of support for different activity-level propositions in the case. We also develop a modification of the standard chain event graph to simplify their use in forensic applications.

stat.AP↗

arXiv2019

Using extreme value theory for the estimation of risk metrics for capacity adequacy assessment

This paper investigates the use of extreme value theory for modelling the distribution of demand-net-of-wind for capacity adequacy assessment. Extreme value theory approaches are well-established and mathematically justified methods for estimating the tails of a distribution and so are ideally suited for problems in capacity adequacy, where normally only the tails of the relevant distributions are significant. The extreme value theory peaks over threshold approach is applied directly to observations of demand-net-of-wind, meaning that no assumption is needed about the nature of any dependence between demand and wind. The methodology is tested on data from Great Britain and compared to two alternative approaches: use of the empirical distribution of demand-net-of-wind and use of a model which assumes independence between demand and wind. Extreme value theory is shown to produce broadly similar estimates of risk metrics to the use of the above empirical distribution but with smaller sampling uncertainty. Estimates of risk metrics differ when the approach assuming independence is used, especially when data across different historical years are pooled, suggesting that assuming independence might result in the over- or under-estimation of risk metrics.

stat.AP↗