The effect of a user selected number of contributors within the LR assignment. Issue 4 (4th July 2022)

• Record Type: Journal Article
• Authors: Kelly, Hannah; Bright, Jo-Anne; Kruijver, Maarten; Taylor, Duncan; Buckleton, John
• Is Part Of: Australian journal of forensic sciences. Volume 54:Issue 4(2022)
• Page Start: 450
• Page End: 463
• Publication Date: 2022-07-04
• Subjects: DNA mixtures -- number of contributors -- forensic DNA analysis -- forensic software -- STRmix -- probabilistic genotyping -- likelihood ratio; Law -- Australia -- Periodicals; Forensic sciences -- Periodicals; Forensic Medicine -- Periodicals; 363.2505
• DOI: 10.1080/00450618.2020.1865456
• Languages: English
• ISSNs: 0045-0618

Abstract

ABSTRACT: The assignment of the number of contributors (N) to a forensic DNA profile is undertaken as part of the interpretation process. There is no requirement for N to be the same for both propositions within the likelihood ratio framework. ISFG recommendations on mixture interpretation suggest that there may be times where prosecution and defence both specify their own, different N. We investigate how this affects the likelihood ratio (LR) for 100 mixed DNA profiles. We show that the addition of a superfluous unknown contributor within a proposition tends to increase the likelihood because the profile can always be better described with more contributors but the priors on the additional parameters and an extra genotype can ultimately reduce the likelihood ratio. Additionally, we find that choosing improbable values for N given different propositions can lead to misleading LRs. Specifically, we find that LRs can both increase and decrease compared to the value obtained using the ground truth N. Choosing a single N which maximizes the probability of the observations for each party tends to approximate an exhaustive stratified LR that takes into account different Ns with different prior probabilities.