One of the Daubert standard’s illustrative factors is a known error rate. Gross errors, such as contamination and sample swapping do happen, as well as many errors with lesser consequences. However, many commentators seek to include false support for non-donors under the classification of error. This appears to be a terminology difference between scientists and lawyers. Scientists would not consider these errors but rather as part of the expected performance of any system with less than 100% discrimination.
Using what we presume is the legal view that false support is an error, again some commentators seek to inform the rate of this using the exceedance probability or the p-value. This is the probability that a false donor would give an LR greater than 1 or greater than the LR in this case, respectively. This erroneously treats the LR as a categorical variable. The exceedance probability systematically underrepresents the weight of evidence, and the p-value can overstate the value of the evidence. The only route to the correct logic is to combine the LR and the prior odds.
(Publisher abstract provided.)