Table 3.
Summary statistics for estimates under a gene-environment interaction model, logit(Pr(D = 1 |G1, G2, G3, E)) = a + b1G1 + b2G2 + b3G3 + b4E + b5G1 × E
| Type | b1 = 0.405 | b2 = 0.405 | b3 = 0.405 | b4 = 0.693 | b5 = 1.100 |
|---|---|---|---|---|---|
| n1 = 250, n2 = 250 | |||||
| Triads | 0.386 (0.295, 0.279, 94.1) | 0.340 (0.214, 0.219, 94.3) | 0.354 (0.224, 0.218, 94.4) | – | 0.834 (0.498, 0.463, 88.9) |
| CC | 0.410 (0.270, 0.263, 94.1) | 0.408 (0.246, 0.231, 93.0) | 0.404 (0.230, 0.230, 95.2) | 0.710 (0.246, 0.247, 95.1) | 1.190 (1.020, 0.559, 97.0) |
| Combined | 0.391 (0.243, 0.231, 94.8) | 0.367 (0.195, 0.189, 93.5) | 0.372 (0.192, 0.189, 94.5) | 0.743 (0.241, 0.240, 94.9) | 0.932 (0.418, 0.400, 91.6) |
| n1 = 375, n2 = 375 | |||||
| Triads | 0.377 (0.231, 0.227, 94.4) | 0.349 (0.174, 0.178, 94.2) | 0.355 (0.184, 0.178, 93.0) | – | 0.822 (0.382, 0.373, 87.5) |
| CC | 0.407 (0.208, 0.214, 95.6) | 0.411 (0.187, 0.187, 94.8) | 0.399 (0.188, 0.187, 94.7) | 0.698 (0.195, 0.200, 95.2) | 1.160 (0.471, 0.451, 95.3) |
| Combined | 0.388 (0.187, 0.188, 94.7) | 0.374 (0.151, 0.154, 94.0) | 0.372 (0.159, 0.154, 93.6) | 0.737 (0.188, 0.196, 95.4) | 0.937 (0.341, 0.326, 91.0) |
| n1 = 500, n2 = 500 | |||||
| Triads | 0.378 (0.198, 0.195, 94.5) | 0.348 (0.150, 0.154, 94.8) | 0.357 (0.159, 0.154, 92.6) | – | 0.813 (0.331, 0.321, 82.9) |
| CC | 0.406 (0.187, 0.185, 94.3) | 0.413 (0.163, 0.162, 95.4) | 0.411 (0.161, 0.162, 95.6) | 0.704 (0.172, 0.173, 95.0) | 1.110 (0.389, 0.384, 95.9) |
| Combined | 0.389 (0.164, 0.163, 94.9) | 0.375 (0.134, 0.133, 94.5) | 0.379 (0.135, 0.133, 93.9) | 0.739 (0.166, 0.169, 94.7) | 0.914 (0.281, 0.280, 89.1) |
The results are based on 1000 simulated datasets, each consisting of n1 case-parent triads and n2 unrelated controls. ‘Triads' refers to using only triad data for estimation; ‘CC’ refers to using only cases (not parents) and unrelated controls; ‘Combined’ refers to using both case-parents and unrelated controls. Each entry lists the mean estimate (standard deviation of the estimates, mean of the estimated standard errors, 95% coverage probability) over the 1000 simulated datasets. – = The main effect of the environmental covariate is unestimable from triads-only.