To test if a self‐control trait can explain the association between CADM2 and risky behavior, we tested mediation following Baron and Kenny's procedure (see Figure 1, including p‐values rather than regression weights as MAGMA does not provide such estimates). 47 We first tested the relationship between CADM2 and the risk behavior factor (path c) in MAGMA, and if that was significant, we tested the association between the self‐control trait (mediator) and the risk behavior factor in SPSS (path b). If path b and path c were significant, and there was an association between a self‐control trait and CADM2 in the gene‐based test (path a), we tested in a final step the relationship between CADM2 and the risk behavior factor outcome, while controlling for the self‐control mediator, in MAGMA (path c'). When in path c' the relationship between the risk behavior and CADM2 was attenuated while controlling for self‐control, mediation was assumed. 48 In all paths, we controlled for the effects of age, sex, and education, and in the analyses involving genetic data, we controlled for the PCs.
