Fig. 3
Δlog10(p-value) between sPLINK and PLINK as well as the set of SNPs identified by sPLINK and PLINK as significant for logistic regression (a, d), linear regression (b, e), and chi-square test (c, f), respectively. For most of the SNPs, the difference is zero, indicating that sPLINK gives the same p-values as PLINK. The negligible difference between p-values for the other SNPs can be attributed to differences in floating point precision. The spikes in some genomic positions are due to the strong association of the corresponding SNPs, which result in higher absolute error. sPLINK and PLINK also recognize the same set of SNPs as significant. Genomic positions (ticks in a–c) indicate chromosome numbers. The details of the experiments are available in Additional file 1: Table S1