Beyond traditional paternity and identification cases. Selecting the most probable pedigree.

摘要

The paper extends on the traditional methodology used to quantify DNA evidence in paternity or identification cases. By extending we imply that there are more than two alternatives to choose between. In a standard paternity case the two competing explanations H(1): "John Doe is the father of the child and H(2): "A random man is the father of the child, are typically considered. A paternity index of 100000 implies that the data is 100000 more likely assuming hypothesis H(1) rather than H(2). If H(2) is replaced by "A brother of John Doe is the father", the LR may change dramatically. The main topic of this paper is to determine the most probable pedigree given a certain set of data including DNA profiles. In the previous example this corresponds to determining the most likely relation between John Doe and the child. Based on DNA obtained from victims of a fire, bodies found in an ancient grave or from individuals seeking to confirm their anticipated family relations, we would like to determine the most probable pedigree. The approach we present provides the possibility to combine non-DNA evidence, say age of individuals, and DNA profiles. The program familias, obtainable as shareware from http://www.nr.no/familias, delivers the probabilities for the various family constellations. More precisely, the information (if any) prior to DNA is combined with the DNA-profiles in a Bayesian manner to deliver the posterior probabilities. We exemplify using the well published Romanov data where the accepted solution emerges among 4536 possibilities considered. Various other applications based on forensic case work are discussed. In addition we have simulated data to resemble an incest case. Since the true family relation is known in this case, we may evaluate the method.