Bayesian meta-analysis with systematic error

R

• get_eumaeus_data.R can be used to extract Eumaeus results to apply this method to (requires database key). Those data should be saved into the data directory (git-ignored).
• get_est_functions.R contains functions for applying stan models to historical controls data
• compare_likelihoods.R compares the timing/results for those models, which use different likelihoods

stan

• metanalysis-likelihood.stan is the stan model that implements the method for an arbitrary likelihood. Requires a set of negative control estimates and standard errors per site and a approximation of the profile likelihood of the effect of interest for each site. The prior distributions are specified here but the parameters in the R code. This model assumes that the likelihood was evaluated along a regular series of points (now forced to be in the R code).
• metanalysis-poisson.stan uses the poisson likelihood (e.g., for historical controls data) to do the same thing. Requires the event counts/person-time for the exposed and comparison group for both the effect of interest and the negative controls (see R code).
• metanalysis-normal.stan uses the normal likelihood for both the effect of interest and the negative controls. Requires point estimates and standard errors for each.

data

• Where the Eumaeus results get saved.

renv

• R packages

documents

results