Table 3.
Multiplex Family Simulations of Family-Biased Ascertainment
| M | Average n | Set | Prevalence | Set | ATT | MLM | CARAT | LTMLM | LT-Fam |
|---|---|---|---|---|---|---|---|---|---|
| 10,000 | 5,000 | unbiased | 10% | causal | 9.462 (0.132) | 8.396 (0.117) | 8.398 (0.117) | 7.174 (0.100) | 8.995 (0.125) |
| null | 1.069 (0.002) | 0.925 (0.001) | 0.925 (0.001) | 0.787 (0.001) | 1.002 (0.001) | ||||
| all | 1.153 (0.002) | 1.000 (0.002) | 1.000 (0.002) | 0.851 (0.002) | 1.082 (0.002) | ||||
| all λGC | 1.083 (0.003) | 0.938 (0.002) | 0.938 (0.002) | 0.797 (0.004) | 1.016 (0.002) | ||||
| 3971 | concord | causal | 4.880 (0.073) | 2.803 (0.042) | 2.704 (0.040) | NA | 3.273 (0.049) | ||
| null | 1.505 (0.002) | 0.957 (0.001) | 0.983 (0.001) | NA | 1.002 (0.001) | ||||
| all | 1.539 (0.002) | 0.975 (0.001) | 1.000 (0.001) | NA | 1.025 (0.002) | ||||
| all λGC | 1.525 (0.004) | 0.967 (0.003) | 0.994 (0.002) | NA | 1.014 (0.002) | ||||
| 51 | discord | causal | 0.737 (0.011) | 0.737 (0.011) | 1.175 (0.017) | 1.454 (0.017) | 1.179 (0.017) | ||
| null | 0.624 (0.001) | 0.624 (0.001) | 0.998 (0.001) | 1.543 (0.002) | 1.001 (0.001) | ||||
| all | 0.625 (0.001) | 0.625 (0.001) | 1.000 (0.001) | 1.542 (0.002) | 1.003 (0.001) | ||||
| all λGC | 0.619 (0.005) | 0.619 (0.005) | 0.983 (0.013) | 2.023 (0.030) | 0.990 (0.006) |
We report average association statistics (S.E. in parentheses) at causal markers, null markers, and all markers, as well as λGC computed using all markers, for each method. NA indicates either lack of convergence or a singular phenotypic covariance matrix. We analyze three types of datasets. Unbiased corresponds to a random sample of 10 individuals taken from a family simulated over four generations repeated 500 times to get 5,000 samples. The concordant set corresponds to downsampling case subjects who do not have a case relative in the dataset and control subjects who have a case relative. The discordant set was constructed by greedily matching case subjects to related control subjects.