Statistical Evaluation of Haploid Genetic Evidence

Abstract

A variety of forensic cases need to be based on haploid DNA. The statistical evaluation of the genetic evidence in such cases requires particular attention and methods. There may be considerable differences between databases and this must be accounted for when there is uncertainty about the population origin of the perpetrator. We here assume access to the relevant databases sampled from populations that the perpetrator may conceivably come from. Intuitively, if there are strong reasons to claim that a specific database is particularly relevant, the likelihood ratio should be influenced accordingly. Moreover, the haploid evidence is typically weaker than for nuclear DNA and there is therefore a greater need to assess uncertainty; here we have chosen the bootstrap method. We also discuss the ability of the proposed method to account for population stratification when computing the LR. A pilot implementation of the haploeve software is freely available.