We assessed the sensitivity of our results to the choice of prior by conducting the same analyses using a much smaller prior (0.022) and much larger prior (202). We found that overall the choice of prior had little effect on the results. Specifically for values we report in the main text, the median credible set sizes were unaffected in all analyses. For the larger prior, the number of single-marker credible sets was unaffected except for analysis B in UK Biobank (from 123 to 122), and the median proportion of markers in the credible set was unaffected in all analyses. For the smaller prior, the number of single-marker credible sets only changed for analysis A, going from 78 to 75 in GIANT, and 85 to 86 in UK Biobank, and the median proportion of markers in the credible set increased slightly in all analyses (maximum increase from 0.047 to 0.051).
