Estimation of a Generalized Linear Mixed-Effects Model with a Finite-Support Randoni-KtTects Distribution via Gibbs Sampling

摘要

We discuss a Huyesian hierarchical lU'ticrali/iMl linear mixed ellecls model with a finite suppoit ran doni eH'ee(s distribution and show how (iihhs sampling clltl be used lot cstiniaiini! ilu- posU'rinr ilisiri butiiin til ilu' p.ii.uiuMi-is ami loi clustciiii)' on llir basis ol KmiMluilin.il data. Wlu*n directly sampluij! I'rtnii tin* coiiililioiial diMiihntiims is laborious, tin* adaplive rejection samplinj'. (AKS. (iiiKs and Wn n, I"1).1,; (iiiKs. I')1),!) al|;oiilhiu or adaptive rejection Metropolis sainplinj; tAKMS. tiiiKs i-t ill.. l'W5i algoitthni is used. Log concavity, a preivquiMtc til' AKS, u( the conditional distributions is I'Mtmimrd. We also discuss a Hayesiau solution to ihv uncfiiaimy ol the support si/e ol ilie lamlom eikvts disiii hulittit in suttisiical inleu'iiec, The nit'lhoitolo|.iy is dhisiraied with an analysis ot ilaia Ironi a study ot rejjulatioti of serum purnlhyioiil hounone secietion.