Some thoughts on the role of probabilistic reasoning in the evaluation of evidence

Abstract

Various aspects of the role of probabilistic reasoning in the evaluation of evidence are described. These include relative frequencies, discriminating power, significance tests and likelihood ratios, and comments on new developments to aid evidence evaluation. The relevance of all these concepts for soil evaluation is considered as appropriate. It is shown that a procedure based on the likelihood ratio emphasises that information from answers to two opposing and relevant questions needs to be considered. It is shown that the likelihood ratio is the factor which converts a prior odds in favour of a prosecution proposition into a posterior odds. The importance of considering evidence at various levels, source, activity and crime, of propositions are discussed. At present, it is not possible to develop models for likelihood ratios in soil analyses in a way that is available for the elemental analyses of glass fragments or the chemical analyses of drug samples. It is shown that the methodology models variability in characteristics in such a way as to account for variation both between source and within source so that the effect on the odds in favour of the ultimate issue can be measured on a continuous scale. Work that is necessary to be done in order to develop likelihood ratios is highlighted together with the difficulties that are particular to soil analyses.