Conversely, if we assume that the average effect is an odds ratio of ORav =1.02 (consistent with the theory of infinitesimal effects, each having an almost imperceptible contribution[14]), then the prior odds of association will be much higher, because a much larger set of (infinitesimally) associated variants are anticipated and a smaller Bayes Factor would be needed to reach a 3:1 posterior. Moreover, large effects emerging from small studies would be incredible, regardless of their p-value, while very small effects emerging with modest p-values from large studies would provide credible evidence for association.
