In our simulation study, the total number of SNPs, , was set to be 20,000, and the sample size of each data set, , was set at 2000, 5000 or 10000, respectively. The number of risk SNPs was the same for the two diseases and was set at 500, 1000 or 2000, respectively. We varied the proportion of shared risk SNPs between the two diseases, , from to 1. Note that corresponds to the absence of pleiotropy. The disease prevalence, , was fixed at 0.1 and the variance explained by all the SNPs, , was fixed at 0.6 for each of the two diseases. Here and were fixed at 0.4 and 0.1, respectively.
