We further explored how our results compare with previously published ones. Using the SNP effect concordance method (Nyholt 2014) and the NTR as a replication sample, we checked whether there is an excess of SNPs showing concordant effects in the meta-analysis by Verweij et al. (2013) and in our analysis. Of the 2 110 385 HapMap SNPs tested in both samples, we selected for the comparison 25 204 independent HapMap SNPs (r2 > 0.1) that showed the most significant association P-values in the meta-analysis sample. Although we compare summary results for the same phenotype (cannabis initiation) such an analysis is similar in scope to a search for significant pleiotropic effects (genetic overlap): we aimed to single out sets of SNPs showing concordant effects in the two samples beyond what is expected by chance. Concordance of effects was assessed by exact binomial tests. We observed no significant excess of SNPs with concordant effects in the two datasets. It is possible that the effects of the causal variants are too small to be accurately captured by the two samples. It is also
