Testing the Psychometric Properties of a Chinese Version of the Level of Expressed Emotion Scale

Table 4

Summary of fit indices of three-hypothesized models of LEE scale (𝑁=262).

Model	𝜒^2	df	𝜒^2/df	𝑃 value	GFI	AGFI	TLI	RMSEA (90% CI)	SRMR	WRMR
Model 1
Uncorrelated factors	98.34	50	1.97	0.80	0.97	0.96	1.00	0.040 (0.036–0.044)	0.039	0.85
Correlated factors^a	92.58	48	1.93	0.75	0.99	0.98	1.02	0.031 (0.027–0.035)	0.028	0.78
Model 2
Uncorrelated factors	102.33	50	2.05	0.54	0.88	0.87	0.89	0.052 (0.044–0.060)	0.050	0.92
Correlated factors	97.02	48	2.02	0.58	0.89	0.90	0.95	0.050 (0.042–0.058)	0.054	0.89
Model 3
Uncorrelated factors	134.21	50	2.68	0.20	0.86	0.85	0.89	0.071 (0.061–0.081)	0.071	0.99
Correlated factors	125.88	48	2.60	0.25	0.89	0.88	0.91	0.067 (0.055–0.076)	0.060	0.94

Model 1: four-factor model identified by Li and Arthur [12] and in this research, Model 2: two-factor model suggested by Startup in 1999 [10], and Model 3: three-factor model suggested by Kim and Miklowitz in 2004 [32].
aModel fit indices tested with paths (correlations) set-up between the hypothesized factors in each model.
𝜒²: Chi-squared goodness of fit; df: degree of freedom; 𝑃 value (a good fit if 𝑃≧0.1); GFI: goodness-of-fit index (ranging from 0 to 1, a good fit if GFI≧0.9); AGFI: adjusted goodness of fit index (similar to GFI, a good fit if AGFI≧0.9); TLI: Tucker-Lewis index (0.90–0.95 acceptable, a good fit if TLI>0.95); RMSEA: root mean square error of approximation (a good fit if RMSEA≦0.05); SRMR: standardized root mean square residual, (a good fit if SRMR<0.05); WRMR: weighted root mean residual (a good fit if WRMR<0.90).