Fig. 1

Comparison of CIMBAL with complete case meta-analysis approach and gold standard (oracle) approach across different simulated data scenarios. The log-odds estimate of the exposure-outcome association (\(\hat {\beta }=\log (\hat {\text {OR}})\)) and its SE \(\left (\hat {\text {se}}=\sqrt {\hat {\text {Var}}(\hat {\beta })} \right)\) from the combined cohort over 2500 independent replicate datasets are plotted for each scenario: (1) fewer cohorts or (2) equal number of cohorts or (3) more cohorts with no confounder information than with complete confounder information. The horizontal dashed line in the \(\hat {\beta }\)-plots correspond to the true β=0