Bayesian Methods in Health Economics

Statistical cost-effectiveness analysis 105

Moreover, it should be noticed that the DIC has a formal justification

for models where the posterior distribution of the parameters of interest is

(approximately) Normal. When the sample size is large enough, asymptotic

arguments can be brought forward, but for models informed by limited evi-

dence this might be a crucial limitation.

Nevertheless, the DIC is a simple tool, which can be easily obtained as a

by-product of MCMC procedures and can be used to produce model-averaged

results in a straightforward manner.

If we consider the set of models M =(M

1

,...,M

H

), we can compute the

value DIC

h

for each of them and ...