Calibration of STRmix LRs following the method of Hannig et al .

Calibration may be used to assess whether methods of LR assignment are reliable. Ramos and Gonzalez-Rodriguez [1] introduce the concept of calibration using weather forecasting as an example. Weather forecasters often give a probability of rain. Let us imagine that we wish to check whether these probabilities are being assigned sensibly. If we can assemble a number of days for which the prediction is, for example, around 50%, and of those days about half have precipitation, then this is evidence that this forecaster is operating sensibly at least in this part of the probability range. This approach for assessing LR calibration is based on assessing the calibration of posterior probabilities for ground-truth known examples with varying prior probabilities. This is based on the LR being the multiplier that converts a prior probability into a posterior probability.