Thirdly, we conducted two meta‐analyses for those phenotypes that were present in multiple datasets in order to maximize power to detect associations. The first meta‐analysis was performed on the results from the per‐cohort gene‐based tests using the meta‐analysis procedure in MAGMA. This method aggregates the Z‐values for the gene‐based associations within the individual cohorts while taking sample size into account, in a procedure similar to “normal” meta‐analysis. The results give an indication of the strength of the association with CADM2 across cohorts. The second meta‐analysis was used to get per‐SNP effects that can be used to estimate the variance in the phenotype explained by SNPs in the gene (R 2). To conduct these meta‐analyses, odds ratios for binary outcome variables were converted to betas with corresponding standard errors in the input files and all continuous variables were standardized. The meta‐analysis was conducted in METAL 44 based on standard errors and effect estimates (rather than on sample size) so that β and se(β) could be obtained.
