Table 6.

Simulation study results based on 1000 replications. “Full” refers to fitting all interaction terms using a linear model; “GlobalB” is the globallboosttest algorithm; “HAIC” is the hereditary constraints AIC model; “Boosting” is the dedicated boosting algorithm.

	Full	AIC	HAIC	BIC	Lasso	GlobalB	Boosting
Non-zero coefficients
Age	1.00	0.46	0.14	0.09	0.09	0.00	0.57
Amount of exercise	1.00	0.49	0.11	0.05	0.47	0.18	0.53
Exercise at age 18	1.00	0.42	0.11	0.02	0.40	0.12	0.44
% Calories from protein	1.00	0.25	0.06	0.01	0.11	0.00	0.28
Education level	1.00	0.50	0.13	0.03	0.22	0.00	0.50
Ever smoking	1.00	0.33	0.08	0.03	0.41	0.10	0.42
Hispanic	1.00	0.16	0.02	0.00	0.29	0.04	0.21
African American	1.00	0.60	0.16	0.11	0.66	0.53	0.66
Asian/Pacific Islander	1.00	0.23	0.06	0.01	0.39	0.13	0.30
American Indian	1.00	0.38	0.01	0.02	0.46	0.19	0.43
Zero coefficients
Exercise at age 35	1.00	0.22	0.04	0.00	0.25	0.03	0.22
Exercise at age 50	1.00	0.17	0.04	0.00	0.28	0.05	0.24
% Calories from carbo.	1.00	0.26	0.03	0.00	0.06	0.00	0.19
% Calories from fat	1.00	0.26	0.03	0.00	0.10	0.00	0.17
Current smoking	1.00	0.17	0.04	0.01	0.32	0.06	0.23
Alcohol	1.00	0.20	0.05	0.00	0.21	0.01	0.24
Region middle	1.00	0.16	0.02	0.00	0.29	0.03	0.23
Region south	1.00	0.15	0.03	0.00	0.26	0.02	0.22
Overall summary
MIaSE	0.0570	0.0532	0.0440	0.0456	0.0380	0.0395	0.0312
True Positive	1.0000	0.3819	0.0879	0.0369	0.3492	0.1293	0.4337
False Positive	1.0000	0.1979	0.0331	0.0033	0.2218	0.0249	0.2172