We could also alternatively choose an improper prior, but BUGS and JAGS don’t play very nicely with them. We could choose a flat prior, but would have to explicitly specify start and stop points. The first level of the hierarchical model assumes \(\). We assume a normal approximation to calculate a log odds ratio (\(y_j\)) and standard error (\(s_j\)) for each study. Mortality ranges from 1% to 17% with sample sizes from 50 to more than 2,300. For that, take a look at this set of notesĬonsider data from 8 RCT’s of magnesium sulfate for myocardial infarction. Much of this material comes from a workshop run by Keith Abrams in Boston in the spring of 2008.īut first, note that I am not here addressing the critically important issues of searching for and evaluating the studies that go into a meta-analysis, or even the (sometimes) contentious issue of meta-analysis itself.
#Bayesian in comprehensive meta analysis how to
The rest of the material on this page goes into details and explains how to conceptualize and code a Bayesian meta-analysis.
Don’t worry if it doesn’t entirely make sense right away (though if it does, kudos). Let’s first go through a quick illustration of a Bayesian meta-analysis. We’ll pick up from the previous section on hierarchical modeling with Bayesian meta-analysis, which lends itself naturally to a hierarchical formulation, with each study an “exchangeable” unit. Bayesian Analysis for Epidemiologists Part IV: Meta-Analysis Introduction: Meta-analysis of Magnesium clinical trials