Table 7.
Average Results from Bayesian Multi-SNP Meta-Analyses of Simulated Data Replicates
| SNP 1 |
SNP 2 |
SNP 3 |
SNP 4 |
SNP 5 |
σHap |
|
|---|---|---|---|---|---|---|
| Number of studies |
7 |
4 |
13 |
1 |
7 |
– |
| True OR | 1 | 1.5a | 1 | 1 | 1 | – |
| Mean posterior probability (SE) | 0.02(0.03) | – | 0.23(0.22) | 0.81(0.21) | 0.02(0.04) | 1.00(0) |
| Mean OR (SE) | 1.00(<0.01) | – | 1.04(0.12) | 1.34(0.12) | 1.00(<0.01) | 0.11(0.02)b |
| Mean BCI length | 0.02 | – | 0.34 | 0.42 | 0.02 | 0.10 |
| Mean OR present (SE)c | 0.98(0.06) | – | 1.19(0.25) | 1.39(0.03) | 0.96(0.05) | 0.11(0.02)b |
| Mean BCI length presentc | 0.18 | – | 0.56 | 0.29 | 0.16 | 0.10 |
Results are means (SE) of estimators over 20 analyses of replicate data sets simulated under identical conditions. In each analysis, OR estimates were taken as the median of the posterior sample.
a
Deleting data on the causal SNP 2 in these replicates simulated a scenario in which the causal SNP is unobserved.
b
These are point estimates of σHap and therefore are not ORs.
c
Calculated on the condition that the SNP was included in a model.