Table S5.

Univariate Liability Threshold Model Results for Tobacco Use

Model	Base	EP	Model Fit	df	AIC	Δ–2LL	Δ df	p	Parameters for Most Parsimonious Model
			–2LL						A (CI)	C (CI)	E (CI)
Sat	–	10	4755.55	4894	–5032.45	–	–	–	–	–	–
Sub 1	Sat	6	4757.02	4898	–5038.98	1.47	4	.650	–	–	–
Sub 2	Sat	4	4758.86	4900	–5041.14	3.31	6	.202	–	–	–
ACEa	Sat	5	4758.86	4901	–5043.14	3.31	7	.860	0.32 (0.17–0.49)	0.51 (0.36–0.64)	0.17 (0.13–0.21)
AE	ACE	4	4792.97	4902	–5011.03	34.12	1	<.001	–	–	–
CE	ACE	4	4776.80	4902	–5027.20	17.94	1	<.001	–	–	–
E	ACE	3	5266.30	4903	–4539.70	507.44	2	<.001	–	–	–

Note: Δdf = difference in degrees of freedom comparing each model to the base model; Δ–2LL=log-likelihood ratio χ² test comparing the –2LL fit of each model to the –2LL fit of the base model; –2LL= minus 2 log-likelihood. A = additive genetic influences; AIC = Akaike’s Information Criterion (lower values reflect a better fit); Base = comparison model (full ACE compared to saturated models and sub models compared to full ACE); C = common environmental influences; E = unique environmental influences; df = degrees of freedom; EP = number of estimated parameters; Sub 1 = Submodel 1 with equal thresholds across twin order; Sub 2 = Submodel 1 with equal thresholds across twin order and zygosity.

^aMost parsimonious model.