Table 6.
Predictions of the predictive accuracy of RSR and of the gain in predictive accuracy of RR compared to RSR in large scale samples (e.g., N ≥ 100,000) based on a linear model with sample size, number of SNPs, fraction of causal SNPs, and heritability as predictors. 95% confidence intervals (CI) are reported in parentheses, with the middle value indicating the point estimate. 1% of the SNPs are assumed to be causal.
| N | P | h SNP 2 | 95% CI R RSR 2 | 95% CI R RR 2/R RSR 2 |
|---|---|---|---|---|
| 100 k | 500 k | 0.50 | (0.139; 0.146; 0.153) | (1.062; 1.070; 1.079) |
| 200 k | 500 k | 0.50 | (0.324; 0.337; 0.349) | (1.094; 1.107; 1.121) |
| 500 k | 500 k | 0.50 | (0.473; 0.478; 0.482) | (1.142; 1.167; 1.193) |
|
| ||||
| 500 k | 100 k | 0.50 | (0.486; 0.488; 0.490) | (1.218; 1.244; 1.270) |
| 500 k | 200 k | 0.50 | (0.482; 0.485; 0.488) | (1.193; 1.218; 1.244) |
| 500 k | 500 k | 0.50 | (0.473; 0.478; 0.482) | (1.142; 1.167; 1.193) |
|
| ||||
| 500 k | 500 k | 0.25 | (0.205; 0.212; 0.218) | (1.110; 1.135; 1.160) |
| 500 k | 500 k | 0.50 | (0.473; 0.478; 0.482) | (1.142; 1.167; 1.193) |
| 500 k | 500 k | 0.75 | (0.733; 0.736; 0.739) | (1.191; 1.218; 1.245) |