Health status transitions in community-living elderly with complex care needs: a latent class approach

BMC Geriatrics

Table 3 Model fit statistics for latent class analysis at baseline, T1, T2

	T0				T1		T2
	2 classes	3 classes	4 classes	5 classes	4 classes	5 classes	4 classes	5 classes
Sequential model comparisons	2 vs. 1 classes	3 vs. 2 classes	4 vs. 3 classes	5 vs. 4 classes	4 vs. 3 classes	5 vs. 4 classes	4 vs. 3 classes	5 vs. 4 classes
LMR LRT
Log-likelihood value (c+1 classes)	13265.31	13595.67	12239.68	12093.26	8366.61	8119.41	4603.94	4449.05
-2 difference in log likelihood	1468.19	2711.98	292.85	107.55	494.40	108.374	309.774	77.03
p value	0.000	0.000	0.0006	0.307	0.000	0.286	0.000	0.198
Adjusted LMR LRT	1458.80	2375.46	290.97	106.86	490.03	107.64	265.99	76.43
p value	0.000	0.000	0.0006	0.309	0.000	0.289	0.000	0.198
Information criterion
AIC	25148.43	24483.26	24192.52	24296.97	16244.82	16348.45	9072.11	9039.11
BIC	25365.96	24493.48	24207.69	24848.37	16258.87	16858.66	9426.66	9483.32
Adjusted BIC	25229.38	24487.13	24198.16	24502.15	16249.34	16512.53	9150.57	9137.40
Entropy	0.814	0.791	0.805	0.765	0.815	0.804	0.830	0.849
Condition number	0.0029	0.309	0.233	0.0012	0.0524	0.0014	0.0301	0.022

LMR-LRT = Lo-Mendell-Rubin Likelihood Ratio Test; AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion.
Condition number = ratio of largest Eigen value to the smallest Eigen value for the Fisher information matrix. Values less than 10E^-09 indicate problem with model identification.

ISSN: 1471-2318