TABLE 3.

Simulation results for sample sizes 200 and 500 (adjusted penalty)

	Scenario
	1	2	3	4	5	6	7	8	9
cadda	1.822	1.700	1.694	2.494	2.456	2.353	3.086	3.014	2.600
caddb	1.993	1.990	1.995	3.280	3.238	3.218	5.200	5.163	5.140
cdoma	0.978	0.987	0.966	1.302	1.240	1.180	1.504	1.280	1.032
cdomb	0.938	0.945	0.943	1.865	1.848	1.865	2.833	2.390	2.755
cepia	1.064	0.940	0.482	1.016	0.806	0.380	0.962	0.630	0.252
cepib	1.745	2.365	3.780	1.685	2.220	3.410	1.625	1.963	3.278
fpa	0.156	0.200	0.228	0.260	0.262	0.280	0.370	0.336	0.346
fpb	0.125	0.150	0.230	0.205	0.205	0.260	0.233	0.218	0.270
pfpa	0.039	0.052	0.068	0.051	0.055	0.067	0.062	0.064	0.082
pfpb	0.026	0.028	0.033	0.029	0.027	0.030	0.024	0.022	0.024
Δca.epia	0.018	0.130	0.158	0.046	0.112	0.100	0.058	0.128	0.100
Δca.epib	0.150	0.040	0.070	0.185	0.035	0.085	0.145	0.110	0.395
Δfpaa	0.07	0.11	0.104	0.082	0.09	0.08	0.066	0.086	0.084
Δfpab	0.065	0.04	0.065	0.035	0.025	0.055	0.05	0.027	0.048
pfpaa	0.055	0.074	0.088	0.065	0.069	0.080	0.071	0.075	0.095
pfpab	0.038	0.034	0.042	0.033	0.030	0.035	0.028	0.025	0.027

Simulation results for the two-step procedure based on formula (4) and the three-step procedure based on formula (5) are shown. Columns contain the average numbers of correctly identified additive (c_add), dominance (c_dom), and epistatic (c_epi) effects as well as the average number of falsely detected effects (fp) and the proportion of false positives (pfp) for the two-step procedure. Δc_a.epi and Δfp_a display average numbers of correctly identified and false-positive epistatic effects that were detected additionally in the third search step on the basis of formula (5); “pfp_a” on the other hand denotes the total proportion of false positives based on the finally selected model at the end of the third search step.

^aResults are for a sample size of 200.

^bResults for a sample of size 500.