Table 3.
MAF | r2 | mâx | max | err |
---|---|---|---|---|
0.1 | 0.5 | 0.595 | 0.588 | 0.006 |
0.1 | 0.7 | 0.780 | 0.781 | 0.001 |
0.1 | 0.9 | 0.951 | 0.954 | 0.003 |
0.2 | 0.5 | 0.566 | 0.564 | 0.002 |
0.2 | 0.7 | 0.761 | 0.760 | 0.002 |
0.2 | 0.9 | 0.940 | 0.938 | 0.001 |
0.3 | 0.5 | 0.561 | 0.557 | 0.005 |
0.3 | 0.7 | 0.751 | 0.751 | 0.001 |
0.3 | 0.9 | 0.933 | 0.935 | 0.002 |
0.4 | 0.5 | 0.549 | 0.552 | 0.003 |
0.4 | 0.7 | 0.751 | 0.747 | 0.004 |
0.4 | 0.9 | 0.930 | 0.931 | 0.001 |
0.5 | 0.5 | 0.547 | 0.550 | 0.003 |
0.5 | 0.7 | 0.749 | 0.746 | 0.003 |
0.5 | 0.9 | 0.931 | 0.931 | 0.000 |
We generated correlated triples of binomial data such that two tag SNPs had the same minor allele frequency and were associated with a third SNP at the same level of r2. 1000 samples of 120 triples were used to estimate the empirical maximum of two SNPs (mˆax), and compared to our analytical formula for the maximum of two SNPs (max). The error (err) is the absolute difference between the empirical and analytical estimates. The inflation of coverage is very significant and suggests that estimates of power based on the best tag for each SNP in the HapMap are overly generous.