Bayesian modeling and model–based decision analysis 213
with a variable SEk=σ
k
adding to a dataset include variables dk and study,
with the following code:
mud=Delta+ui;
random uk ~norma(0,sigmau) subject=study;
model dk ~normal(mud,var=SEk**2);
proc MCMC uses the s ame syntax, so this method can also be used in a
Bayesian approach. Bayesian approaches are commonly used in meta analysis
for their better small sample properties, since the number o f studies in a
meta analysis is often small. Trea tment heterogeneity u
k
is the key component
in meta analysis, as it determines if treatment effects can be generalized to
different studies, which are often considered as r epresentatives o f different
populations. The ideal situation is when σ
u
= 0, i.e., treatmen ...