Table 4.

Model fit comparison tests, with the preferred model printed in bold.

Model	Δdf	ΔLL	AIC	p value
ADE			105451.9
Causal ADE	4	155.8	105599.7	<0.001
ADE			105451.9
Causal ADE + A confounding	2	0.407	105448.3	0.815
ADE			105451.9
Causal ADE + E confounding	2	24.8	105472.7	<0.001
ADE			105451.9
Causal ADE + D confounding	2	38.4	105486.3	<0.001

Δdf is the difference of degrees of freedom of the models, ΔLL is the difference in the minus 2 log-likelihood of the models, AIC is the Akaike information criterion. The preferred model is printed in bold.