Table 2.

Fit statistics for global cognitive function group trajectories in middle-aged and older adults from CHARLS

Fit statistic	Number of classes
	1	2	3	4	5	6
BIC*	− 35,903.71	− 32,658.89	− 32,120.03	− 32,098.63	− 32,099.94	− 32,116.00
AIC*	− 35,886.18	− 32,623.83	− 32,074.44	− 32,039.02	− 32,029.82	− 32,031.85
Class proportion¶	Class 1, 100%	Class 1, 38.29%	Class 1, 18.89%	Class 1, 12.49%	Class 1, 13.36%	Class 1, 13.24%
		Class 2, 61.71%	Class 2, 38.93%	Class 2, 21.54%	Class 2, 18.36%	Class 2, 17.61%
			Class 3, 42.17%	Class 3, 31.55%	Class 3, 20.40%	Class 3, 19.34%
				Class 4, 34.42%	Class 4, 36.26%	Class 4, 0.40%
					Class 5, 11.62%	Class 5, 36.18%
						Class 6, 13.23%
APP‡		Class 1, 0.92	Class 1, 0.87	Class 1, 0.80	Class 1, 0.79	Class 1, 0.79
		Class 2, 0.94	Class 2, 0.80	Class 2, 0.67	Class 2, 0.56	Class 2, 0.56
			Class 3, 0.89	Class 3, 0.67	Class 3, 0.50	Class 3, 0.48
				Class 4, 0.84	Class 4, 0.79	Class 4, 0.58
					Class 5, 0.48	Class 5, 0.67
						Class 6, 0.47

CHARLS China Health and Retirement Longitudinal Study, AIC Akaike’s information criterion, BIC Bayesian information criteria, APP average posterior probabilities

*A lower absolute value suggests a better model fit

^¶No less than 5% of total count in a class

^‡A higher value is better (preferably > 0.7 in a class)