简介:Aweightofevidenceisacalibratedstatisticwhosevaluesin[0,1]indicatethedegreeofagreementbetweenthedataandeitheroftwohypothesis,onebeingtreatedasthenull(H0)andtheotherasthealternative(H1).Avalueofzeromeansperfectagreementwiththenull,whereasavalueofonemeansperfectagreementwiththealternative.Theoptimalityweconsiderisminimalmeansquarederror(MSE)underthealternativewhilekeepingtheMSEunderthenullbelowafixedbound.Thispaperstudiessuchstatisticsfromaconditionalpointofview,inparticularforlocationandscalemodels.