Likelihood ratio distributions and the (ir)relevance of error rates

Streamlining the decision-making process for international DNA kinship matching using Worldwide allele frequencies and tailored cutoff log10LR thresholds

“…It is wise to keep in mind that the only purpose of using cutoff log 10 LR thresholds is to retrieve promising items through a database search. The obtained log 10 LR values cannot be extrapolated and used to the probability that H1 (the MP and UHR sample are the same individual) is true [60]. This log 10 LR result…”

“…The goal of a successful DNA database search is to then communicate the evidential value, when it exceeds a certain strength; i.e. the said strength is low enough to enable identifications but high enough to prevent reporting evidence that is coincidental [60]. Indeed, too many potential candidate matches would severely increase the workload of the DNA experts and police officers as many of the potential candidate matches are actually adventitious [61].…”

The analogy between DNA kinship and DNA mixture evaluation, with applications for the interpretation of likelihood ratios produced by possibly imperfect models

“…However, since the distribution of the likelihood ratio depends on which hypothesis is true, we can compute error rates that tell us the converse, predictive probabilities, namely how likely we are to take which decision if either hypothesis is true (cf. [12]).…”

Making decisions in missing person identification cases with low statistical power

“…Two measures of statistical power are particularly useful in large-scale cases such as the "Missing Grand-children of Argentina" [1,4]: The inclusion power (IP) is the probability that LR ≥ T if POI is MP, while the exclusion power (EP) is the probability that a random unrelated person can be excluded due to genetic inconsistencies with the relatives of the missing person (mutation model is not used when estimating EP). Low statistical power increases the risk of not reaching the correct decision [5] (Figure 1). Potential matches, indicated by LR ≥ T , are analysed further, with more data (mtDNA, Y-STRs or X-chromosomal markers), to reach a conclusion about the case.…”

“…These approaches analyzed different types of relationships (uncle-nephew, grandparent-grandchildren, etc) in order to determine LR thresholds for declaring an identification. As mentioned by Slooten et al [5] a distinction should be done between notions of evidence and decision making in these cases. We do not provide a LR threshold for reaching a conclusion in the identification process, when comparing H1: POI is MP, and H2: POI is unrelated to MP.…”