Generalized fiducial factor: An alternative to the Bayes factor for forensic identification of source problems

One formulation of forensic identification source problems is to determine the trace evidence, for instance, glass fragments found on a suspect crime. The current state science compute Bayes factor comparing marginal distribution measurements evidence under two competing propositions whether or not unknown originated from specific source. obvious problem with such an approach ability tailor prior distributions (placed features/parameters statistical model evidence) in favor defense prosecution which further complicated by fact that typical number typically sufficiently small choice/specification has strong influence value factor. To remedy this specification and choice, we develop alternative factor, within framework generalized fiducial inference, term Furthermore, demonstrate empirically, synthetic real Netherlands Forensic Institute casework data, deficiencies classical/frequentist likelihood ratio approaches.