Bayesian Networks for Probabilistic Inference and Decision Analysis in Forensic Science, 2nd Edition

Chapter 7Evaluation of DNA profiling results

Interest in the probabilistic evaluation of DNA profiling results has grown considerably during the past 20 years. Topics such as the assessment of transferred DNA traces in the form of mixtures, the consideration of error probabilities and the effects of database searches have been responsible for the increase in interest in forensic statistics amongst forensic scientists. There are other complications in DNA analysis such as the possibility of the occurrence of a mutation, which have to be considered in certain applications, typically kinship analysis. There are many aspects to the probabilistic evaluation of DNA profiling results, which is reflected in the size of the published literature on the use of Bayesian networks, which consider these evaluative topics. DNA profiling analyses are, arguably, the most reported applications of Bayesian networks. As will become apparent throughout this chapter, an important aspect of Bayesian network modelling for DNA profiles relies on the proper representation of basic entities, such as genes and genotypes, which may be made at different levels of detail. Such principal network fragments may be repeatedly invoked, given recognized and extensive biological theory (e.g. Mendelian inheritance), to represent the genetic constitution of various observed and unobserved individuals. This explains in part why rigorous Bayesian network models may be constructed even for scenarios that are large and ...