Table 2.
Empirical type I error estimates under the null model of a time‐to‐event primary phenotype
| Scenario | Censoring | CIEE | BS | G‐EST | MR |
|---|---|---|---|---|---|
| 1 | 10% | 5.29% | 5.29% | 22.81% | 4.82% |
| 30% | 5.24% | 5.13% | 24.98% | 5.00% | |
| 50% | 5.29% | 5.33% | 20.24% | 5.28% | |
| 2 | 10% | 5.15% | 5.45% | 34.48% | 5.28% |
| 30% | 5.13% | 5.29% | 37.83% | 5.15% | |
| 50% | 5.14% | 5.20% | 30.33% | 4.74% | |
| 3 | 10% | 5.10% | 5.12% | 34.54% | 5.34% |
| 30% | 4.94% | 4.92% | 37.25% | 5.30% | |
| 50% | 4.88% | 4.77% | 30.66% | 4.84% | |
| 4 | 10% | 5.23% | 5.19% | 31.59% | 6.07% |
| 30% | 5.15% | 5.15% | 35.40% | 6.17% | |
| 50% | 5.24% | 5.14% | 29.43% | 5.68% | |
| 5 | 10% | 5.15% | 5.27% | 4.94% | 6.17% |
| 30% | 4.98% | 5.08% | 4.80% | 5.79% | |
| 50% | 4.93% | 4.84% | 4.33% | 5.73% |
Data were generated for individuals and replicates. The MAF of the marker X was set to 0.2. CIEE is the proposed method using estimating equations; BS is CIEE using nonparametric bootstrap standard errors; G‐EST is the sequential G‐estimation approach (Lipman et al., 2011); and MR is multiple log‐linear censored regression.