Figure 1.
Conditional genotype and score distributions.
(A): Conditional genotype distribution; (B): Conditional score distribution without covariate adjustment; and (C): Conditional score distribution with covariate adjustment. The parameters used in our simulations under the balanced case-control design are: = 2000, MAF = 0.5,
= -5.30,
= 2.5,
= 1, and
.
Figure 2.
Asymptotic trends of testing accuracy with different sample sizes.
The result was based on a checkerboard model whose parameters were the same as shown in Figure 1. The solid lines are the analytical accuracy and represent the upper bound of the testing accuracy. The three lines downward are the means of the testing accuracies from 200 simulations with a sample size of 2000, 1000, and 500. Because the lines for a sample size of>2000 are coincident with the analytical accuracy lines, they are not shown.
Figure 3.
Linear correlation between accuracy and heritability.
The solid line in each panel is fitted with the method of least squares, and its is shown in bold font. The dashed lines in 3ULM and 4ULM panels were fitted alike while excluding dots below 0.52, and their
values are shown above the reference lines indicating accuracy of 0.52. For the six regression models, the p value for F test was<0.001.
Table 1.
Theoretical accuracies for the three simulated modelsa.
Figure 4.
Power comparison of GMDR and MDR for sample sizes of 500, 1000, and 2000 under the checkerboard (digenic) model at alpha = 0.05.
For each panel, 12 combinations, as defined in Table 1, were simulated, forming three levels of MAF (0.1, 0.25, and 0.5) and four levels of interactive effects (1.0, 1.5, 2.0, and 2.5). Simulation results from sample sizes of 4000 and 10,000 are not shown because no difference in power estimates were detected by the GMDR and MDR methods.
Figure 5.
Power comparison of GMDR and MDR for sample sizes of 500, 1000, 2000, and 4000 under the 3ULM (trigenic model) at alpha = 0.05.
For each panel, 12 combinations, as defined in Table 1 were simulated, as shown here, which were formed of three levels of MAFs (0.1, 0.25, and 0.5) and four levels of interaction effects (1.0, 1.5, 2.0, and 2.5).
Figure 6.
Power comparison of GMDR and MDR for sample sizes of 500, 1000, 2000, and 4000 under the 4ULM (tetragenic model) at alpha = 0.05.
For each panel, 12 combinations, as defined in Table 1, were simulated, as shown here, which were formed of three levels of MAFs (0.1, 0.25, and 0.5) and four levels of interaction effects (1.0, 1.5, 2.0, and 2.5). Simulation results from the sample of 10,000 are not shown because no difference in power estimates was detected for the GMDR and MDR methods.