Measurement of the accuracy (percentage of the top SNPs correctly returned) of the PrivSTRAT algorithms for selecting top SNPs with mret (the number of returned SNPs) equal to 3 and 5 for the (a) rheumatoid arthritis GWAS dataset and (b) our simulated dataset, with varying values of the privacy parameter, ε. In all four cases, the accuracy increases as privacy decreases (as ε increases). More importantly, in three of the four cases, high accuracy is achieved for reasonable choice of privacy (ε around 2), which should increase with sample size. These results are averaged over 20 iterations.