Comments arising from WJ Thompson "Uncertainty in probabilistic genotyping of low template DNA A case study comparing STRmix and TrueAllele"

Thompson reports a comparison of data from STRmix and TrueAllele. The data he has arises from different inputs to the two software. If the input data are made more similar the outputs become more similar. Thompson argues that the Analytical Threshold, AT, should be varied in casework. This produced different LRs but the analyst would be left deciding what to do with these options. This cannot be based on the LRs but should be based on whether any movement in the AT adds reliable or unreliable data. This is how most laboratories set their AT in the first place. Hence it is pointless, and potentially dangerous, to experimentally vary the AT in casework. The profile is low level and shows at most three peaks. Thompson argues that LR results assuming that the number of contributors (NoC) is 2 or 3 should be reported. Uncertainty in NoC should be treated as a nuisance variable and summed out.