Table 2.

Model fit statistics to determine pathway analysis model selection

	Pathway	Akaike information criteria	Bayesian information criteria	Improvement in model fit compared to demographically adjusted base model	Change in fit (2–1)
(1)	Aβ42 → Aβ40	10,339.25	10,388.24	−180.37	19.62
(2)	Aβ40 → Aβ42	10,358.87	10,407.86	−160.75
(1)	Aβ42 → T	10,511.5	10,560.49	−8.12	6.03
(2)	T → Aβ42	10,517.53	10,566.52	−2.09
(1)	Aβ42 → N	10,475.18	10,524.16	−44.44	33.23
(2)	N → Aβ42	10,508.41	10,557.4	−11.21
(1)	Aβ40 → T	10,518.64	10,567.63	−0.98	−9.49
(2)	T → Aβ40	10,509.15	10,558.13	−10.47
(1)	Aβ40 → N	10,467.95	10,516.93	−51.67	29.41
(2)	N → Aβ40	10,497.36	10,546.35	−22.26
(1)	T → N	10,519.85	10,568.83	0.23	1.67
(2)	N → T	10,521.52	10,570.51	1.9

The smallest Akaike/Bayesian information criteria, denoted in bold, indicate the best-fitting directional pathway. All models adjusted for associations between age and all outcomes