Abstract:
|
Recent recommendations concerning the use of error rates in forensic science have started to shift the focus of evidence interpretation methods away from the subjective Bayesian approach long advocated by the research community. In response to these recommendations, many forensic disciplines have proposed studies similar to the successful black-box and white-box studies in latent print analysis. These types of studies report an average error rate across a population of examiners for a given set of tasks related to identification of source problems. Although this is not the intention, these studies are used to justify the conclusions that an examiner has made in a specific case. As statisticians focused on the identification of specific source problems, it is our view that it is unclear what these studies imply about a specific source identification problem. In this presentation, we will work within the classical paradigm for evidence interpretation based on conditional match probabilities, a type of forensic error rate. We will then relate these to the error rates common in individuality studies and propose a paradigm for evidence interpretation based on forensic error rates.
|