. 2013 Sep 12;33(5):881–899. doi: 10.1002/sim.5963

Table V.

Average df in each dimension for the best fitting models selected through AIC and BIC (left part), and empirical rejection rate for the AIC and BIC-based selection for the hypotheses of linearity and constant risk (right part) for the nine scenarios of exposure–lag–response associations. Results from m = 500 simulated data sets with n_s = 400 subjects

	Average df				Empirical rejection rate

	f(x)		w(ℓ)		H₀ : f(x) = x		H₀ : w(ℓ) = c

f(x) − w(ℓ)	AIC	BIC	AIC	BIC	AIC	BIC	AIC	BIC
Linear-constant	1.50	1.03	1.57	1.02	0.29^*	0.03^*	0.23^*	0.01^*
Linear-decay	1.26	1.00	3.60	3.17	0.18^*	0.00^*	1.00	1.00
Linear-peak	1.22	1.00	4.02	3.72	0.15^*	0.00^*	1.00	0.98
Plateau-constant	2.26	1.54	1.47	1.00	0.82	0.47	0.19^*	0.00^*
Plateau-decay	2.53	1.55	3.49	3.10	0.97	0.54	1.00	1.00
Plateau-peak	2.18	1.21	4.01	3.56	0.85	0.19	1.00	0.93
Exponential-onstant	2.20	1.56	1.43	1.00	0.83	0.52	0.16^*	0.00^*
Exponential-decay	2.36	1.81	3.58	3.12	0.99	0.80	1.00	1.00
Exponential-peak	2.15	1.29	4.05	3.69	0.90	0.27	1.00	0.93

H₀ is true