Table 2.
Simulation Study 1 summary statistics for the genome‐wide bootstrap, conditional likelihood, and single‐SNP bootstrap estimates (significance threshold P < 5 × 10−5)
| Mean bias in logHR estimates | ||||||||
|---|---|---|---|---|---|---|---|---|
| SNP MAF | Data‐generated logHR | Empirical power (%) | Mean fitted logHR | Uncorrected naïve | Genome‐wide bootstrap | Conditional likelihood | Single‐SNP bootstrap | No. of datasets selected |
| 10.3 | 0.50 | 99.5 | 0.50 | 0.00 | −0.07 | 0.00 | 0.00 | 4,974 |
| 34.9 | 0.19 | 77.3 | 0.22 | 0.01 | −0.04 | 0.01 | 0.00 | 3,863 |
| 26.7a | 0.22 | 65.5 | 0.20 | 0.03 | −0.03 | 0.02 | 0.01 | 3,275 |
| 17.1 | 0.17 | 55.7 | 0.26 | 0.04 | −0.03 | 0.03 | 0.01 | 2,785 |
| 15.0 | 0.13 | 52.2 | 0.27 | 0.04 | −0.03 | 0.03 | 0.02 | 2,611 |
| 13.7a | 0.23 | 50.0 | 0.22 | 0.04 | −0.03 | 0.03 | 0.02 | 2,499 |
| 48.8a | 0.27 | 17.2 | 0.14 | 0.05 | 0.00 | 0.04 | 0.03 | 861 |
| 39.0a | 0.21 | 10.0 | 0.12 | 0.07 | 0.01 | 0.06 | 0.04 | 500 |
| 5.8a | 0.19 | 9.3 | 0.21 | 0.13 | 0.02 | 0.12 | 0.09 | 463 |
| 15.8a | 0.09 | 8.9 | 0.15 | 0.09 | 0.01 | 0.08 | 0.06 | 447 |
| 15.8a | 0.13 | 8.4 | 0.15 | 0.09 | 0.01 | 0.08 | 0.06 | 418 |
| 14.6 | 0.14 | 0.7 | 0.12 | 0.16 | 0.08 | 0.15 | 0.13 | 36 |
| 36.3 | 0.14 | 0.6 | 0.07 | 0.12 | 0.07 | 0.11 | 0.10 | 29 |
| 6.8 | 0.27 | 0.2 | 0.16 | 0.26 | 0.15 | 0.23 | 0.19 | 11 |
| 6.9 | 0.12 | 0 | 0.09 | NA | NA | NA | NA | 0 |
a
Minor allele is the risk allele, otherwise major allele is associated with risk.
Comparison with the naïve Cox PH estimates for 15 SNPs generated to have association with time to severe nephropathy in a sample of 5,444 individuals. The rows are ordered by empirical power, which is the proportion of simulated datasets in which the SNP was detected as significant out of 5,000 replications. Mean bias is calculated as the difference between the mean fitted logHR in all datasets and the mean logHR in selected datasets.