Assessing adolescents’ perceived proficiency in critically evaluating nutrition information

Table 3 Model evaluation by goodness-of-fit indices (GOFI) for the a priori specified measurement model M1 in Fig. 2 and the re-specified and data-driven post hoc modified measurement model M2 in Fig. 3 (robust maximum likelihood estimation using the statistical package LISREL)

Model (M) and GOFI goodness-of-fit target value	Absolute GOFI		Parsimony-adjusted GOFI		Incremental GOFI
Model (M) and GOFI goodness-of-fit target value	SB scaled χ² with p-value	SRMR	RMSEA (90% CI)	CFit	CFI	NNFI
M1 (df = 5, N = 1485)	28.83, p = .0000	0.021	0.115 (0.096; 0.135)	0.000	0.995	0.990
M2 (df = 3, N = 1485)	4.34, p = .2272	0.009	0.054 (0.030; 0.081)	0.358	1.000	0.999
target value	p > .05	< .05	< .06 (< .05; < .08)	> .05	> .95	> .95

Note: M2 is the more restricted and nested model obtained from M1 by the addition of the covariance between the uniqueness variance components of items CNL1 and CNL4, and items CNL2 and CNL5
df = degrees of freedom, N = effective sample size (list wise deletion). Goodness-of-fit indices (GOFI) are classified as absolute, parsimony-adjusted and incremental: SB scaled χ² = Satorra-Bentler scaled chi-square, SRMR = Standardized Root Mean Square Residual, RMSEA = Root Mean Square Error of Approximation, CFit = p-value for test of Close Fit (i.e., the probability that RMSEA < 0.05), CFI = Comparative Fit Index, NNFI = Non-Normed Fit Index = TLI = Tucker & Lewis fit index. Bold values imply mediocre to poor data-model fit (the SB scaled χ² p-value for M1 is insignificant owing to large sample size)

ISSN: 1479-5868