
198 Exposure-Response Modeling: Methods and Practical Implementation
Sigma3 0.00937 0.0206 -0.00422 0.00849 0.0215
Sigma4 0.0631 0.0187 0.0504 0.0594 0.0721
var_y 16.5058 1.9074 15.1683 16.2539 17.6840
sigu 22.8149 5.4977 18.9758 21.6388 26.1234
However, diagnostic plots (not shown here) suggested that not all the Markov
chains were properly c onverged. Nevertheless, comparing the Bayesian poste-
rior means of the parameters with the estimates using proc NLMIXED, there
is no substantial difference. The proc MCMC run with the setting above took
about ten times CPU time as that for the proc NLMIXED run. Hence one
may ask what is the added value for the