Bayesian Analysis of DNA Profiling Data in Forensic Identification Applications

Abstract

The utilization of DNA evidence in cases of forensic identi®cation has become widespread over the last few years. The strength of this evidence against an individual standing trial is typically presented in court in the form of a likelihood ratio (LR) or its reciprocal (the pro®le match probability). The value of this LR will vary according to the nature of the genetic relationship between the accused and other possible perpetrators of the crime in the population. This paper develops ideas and methods for analysing data and evaluating LRs when the evidence is based on short tandem repeat pro®les, with special emphasis placed on a Bayesian approach. These are then applied in the context of a particular quadruplex pro®ling system used for routine case-work by the UK Forensic Science Service.

Voluntary Sleep Choice and Its Effects on Bayesian Decisions

“…An important implication of this result is that commonly experienced levels of sleep restriction may place one at a disadvantage if new information is particularly important. Consider the value of new evidence in a courtroom trial setting (Foreman, et al, 1997) or in medical diagnosis (Kononenko, 1993). Sleep deprived individuals may undervalue new information relative to its proper weighting, which may lead to an incorrect verdict or a suboptimal medical course of action.…”

Interpreting DNA Evidence: A Review

“…This is achieved through Bayesian estimation of products of allele proportions using a binomial or trinomial likelihood and uniform priors. Foreman et al (1997Foreman et al ( , 1999) provide a similar argument using a more general multinomial likelihood for the entire vector of allele proportions. A recent analysis of various sampling allowances (Curran et al, 2002) showed that Balding & Nichols's proposed sampling adjustment provides a good approximation for reasonable-sized databases.…”

“…A new Bayesian approach for examining STR databases was developed by Foreman et al (1997Foreman et al ( , 1999. Estimates of F st were obtained for the 3 major UK ethnic groups using FSS quad frequency databases and in the absence of any specific subpopulation data.…”

Forensic identification with imperfect evidence

“…Then Assumption 1 requires that, while the population as a whole need not be homogeneous with regard to racial make-up, if we restrict attention to individuals of a given race, the distribution of DNA profiles is homogeneous within that racial group. For analyses of DNA evidence with non-homogeneous populations see Balding & Nichols (1995), Roeder et al (1998), Foreman, Smith & Evett (1997. We require pr(Z = z\ Y = y, x a = £), the probability that an individual has the observed profile, conditional on his racial group and the data on race and profile for the measured individuals.…”