Metropolis-Hastings vs Gibbs sampling in PKBugs

1 messages 1 people Latest: Jul 17, 2003

Metropolis-Hastings vs Gibbs sampling in PKBugs

From: Yaning Wang Date: July 17, 2003 technical
From: Yaning Wang <yaning@ufl.edu> Subject: [NMusers] Metropolis-Hastings vs Gibbs sampling in PKBugs Date: 7/17/2003 1:24 AM Dear NONMEM users: I know this list may not be the right place to ask this question. But I hope someone may give me some hints. In the example 1 provided by PKBugs interface, some parameters are sampled using Metropolis-Hastings algorithm while other parameters are updated using Gibbs sampler. Could someone explain to me why this is necessary (different algorithms for different parameters)? Thanks for any response in advance. Yaning Wang FYI From PKBugs user guide 4.1.7 Analysis 4.1.7.1 Monitoring "During the analysis the parameters tau, mu, and omega-1 are updated using Gibbs sampling (Geman & Geman, 1984) whereas components of theta are sampled using a Metropolis-Hastings algorithm (Metropolis et al., 1953; Hastings, 1970). " _______________________________________________________