Having the observed -log10 (p-value) along the diagonal line in the Q-Q plots doesn’t mean these tests agree with each other. To investigate the concordance among these three tests, we rank the -log10 (p-value) from the dTDT (or MH test) and pick the top 100 signals. Then we plot these values with their corresponding -log10 (p-value) from the other two tests ( Figure 8 . To dTDT the other two are Case-Control and MH test; to MH test the other two are Case-Control and dTDT). From the figure, we can see that the concordance performs poorly among these tests. The top signals in either dTDT or MH test do not appear in the top list from the rest two tests as expected. There can be several reasons for this discordance: (1) genotyping errors in both linkage and association data; (2) inaccuracy of the imputation; (3) interference from population stratification. The first two issues may be addressed through experimental validation as discussed below. The last issue requires additional design for the tests that we will discuss in Conclusions.
