Table 2.
True R2 for the two arrays with the largest sample sizes (chr20)
| Illumina panel | MAC/MAF | Number of non-NA-R2 variantsa | Mean true R2 | Median true R2 | Total number of variants |
|---|---|---|---|---|---|
| 610-Quad (n = 1992) | MAC <10 | 311,625 | 0.93 | 1.00 | 377,397 |
| MAF <0.5% | 440,489 | 0.93 | 1.00 | 508,198 | |
| MAF <0.5%–5% | 85,270 | 0.93 | 0.96 | 85,278 | |
| MAF >5% | 120,991 | 0.98 | 1.00 | 120,998 | |
| 660W-set1 (n = 941) | MAC <10 | 229,286 | 0.96 | 1.00 | 299,329 |
| MAF <0.5% | 356,643 | 0.95 | 1.00 | 430,073 | |
| MAF <0.5%–5% | 85,195 | 0.94 | 0.97 | 85,201 | |
| MAF >5% | 121,013 | 0.98 | 1.00 | 121,019 |
MAC, minor allele count; MAF, minor allele frequency.
a
NA true R2 emerged owing to being monomorphic (either true or imputed). Some variants may be monomorphic in the 1992 subset, but not in the 3840 samples. The Pearson correlation between a constant and a vector is not defined.