I've been wondering how the commercial outfits and the military test primers. So far haven't found any paper documenting it. Other papers usually assume the failures are randomly distributed about a mean value and use various statistical tricks to estimate minimum and maximum probabilities (Dixon, Bruceton, Probit, Logit, etc.). In Frost's book "Ammunition Making" he uses a weighted distribution but has no reference as to what its derived from. Often data are collected using a staircase or up-down technique to cluster data around 50% probability point to minimize extraneous data points. With the advent of computers there is software at
http://www.neyersoftware.com/ that does iterative fitting. Anybody have any insights into how the DOD or commercial outfits do their testing?