Author(s): Mark W Perlin
Statistical computing can provide an accurate match statistic for forensic identification. The resulting likelihood ratio (LR) quantifies the probative force of evidence, capturing in a single number the strength of match. But the LR may be difficult to explain to a non-statistician. Nor does the LR convey the chance of error, often a juror’s foremost concern. Error can be expressed as a false match probability (FMP). With biological evidence, a false match occurs when someone’s DNA is not present, but has a match statistic at least as large as the reported LR. FMP is the chance of this misidentification occurring. This article shows how to rapidly and accurately calculate the FMP through logarithmic convolution of independent tests. The approach permits FMP evaluation on very large sets, and provides sharper error estimates than the guaranteed 1/LR upper bound. Mathematical theory is presented, along with a DNA case example of sexual assault and database search. By reporting exact error rates on specific evidence data, FMP assists investigators, scientists, lawyers, jurors and judges in their forensic decision-making.