Pre-Experimental Assessment of Forensic DNA Identification Systems

In this thesis we want to produce a methodology to evaluate a kinship identification system, i.e. the set of models and data used to ascertain the identity of an individual, a probabilistic tool devoted to obtain the likelihood ratio supporting (or contradicting) the hypothesis that an individual, the candidate for identification, is a specific member of a family, conditional on the available familial DNA evidence. The thesis considers the likelihood ratio as a random variable and focuses on the evaluation of the probability that a candidate for identification would be correctly classified exploiting the likelihood ratio distributions conditional on each hypothesis. The aim of this work is thus to show how it is possible to make statements about the goodness of an identification system and to demonstrate how this can be applied in a great variety of cases. As secondary objective, we want to show how it is possible to obtain the distributions for the likelihood ratio, finding efficient computational methods to cope with the their huge state space. The proposed system evaluation is specific for each case, does not require any additional laboratory costs, and should be carried out before the identification trial is performed. In a pre-experimental perspective, we want to evaluate whether a system fulfils the requirements of the parties involved. A further objective is to consider and find a solution for some complicating issues affecting the estimation of mutation rates for STR markers.