Table 3.
Empirical and simulation results for P-value thresholds
| Trait | Threshold (P-value) | Empirical |
Simulation |
||||||
|---|---|---|---|---|---|---|---|---|---|
| Positives |
eFDRb | Δ in #of sig. loci (% true) | Positives |
eFDR | Δ in #of sig. loci (% true) | ||||
| False | Truea | False | True | ||||||
| HDL (ntest = 19,840 ntruth = 188,577) |
5 × 10−8 | 1 | 16 | 5.9% | – | 0.28 |
9.5 | 2.9% | – |
| 5 × 10−7 | 1 | 18 | 5.3% | +2 (100%) | 0.89 |
12 | 6.7% | +3.6 (83%) | |
| 5 × 10−6 | 8 | 21 | 28% | +10 (30%) | 4.2 |
17 | 20% | +8.2 (60%) | |
| LDL (ntest = 19,840 ntruth = 188,577) |
5 × 10−8 | 0 | 14 | 0% | – | 0.19 |
13 | 1.5% | – |
| 5 × 10−7 | 3 | 16 | 16% | +5 (40%) | 0.71 |
16 | 4.2% | +3.9 (87%) | |
| 5 × 10−6 | 10 | 19 | 34% | +10 (30%) | 4.6 |
22 | 17% | +9.1 (58%) | |
| TG (ntest = 19,840 ntruth = 188,577) |
5 × 10−8 | 1 | 8 | 11% | – | 0.11 |
9.0 | 1.2% | – |
| 5 × 10−7 | 2 | 10 | 17% | +3 (67%) | 0.54 |
10 | 4.9% | +1.9 (77%) | |
| 5 × 10−6 | 6 | 15 | 29% | +9 (56%) | 3.8 |
13 | 22% | +6.1 (47%) | |
| Height (ntest = 133,653 ntruth = 693,529) |
5 × 10−8 | 0 | 157 | 0% | – | 1.6 |
181 | 0.89% | – |
| 5 × 10−7 | 1 | 217 | 0.46% | +61 (98%) | 4.9 |
223 | 2.2% | +46 (93%) | |
| 5 × 10−6 | 2 | 312 | 0.64% | +96 (99%) | 16 |
283 | 5.4% | +72 (84%) | |
| BMI (ntest = 123,865 ntruth = 681,275) |
5 × 10−8 | 0 | 22 | 0% | – | 0.62 |
39 | 1.6% | – |
| 5 × 10−7 | 0 | 37 | 0% | +15 (100%) | 2.7 |
58 | 4.4% | +22 (90%) | |
| 5 × 10−6 | 1 | 55 | 1.8% | +19 (95%) | 11 |
90 | 11% | +41 (79%) | |
a
Number of loci in truth set for HDL: 89, LDL: 72, TG: 60, height: 1100, BMI: 724.
b
eFDR is calculated as number of false positives divided by sum of true and false positives.