Current Gerontology and Geriatrics Research

Research Article

What the CERAD Battery Can Tell Us about Executive Function as a Higher-Order Cognitive Faculty

Table 2

Model-data fit two multifactor and one single-factor model of CERAD EF-type tests. Results are shown by study sample and according to whether the total MMSE score, the WORLD backwards item and remaining MMSE total score, or just the WORLD backwards item were included.


Model	Group	Fit Criteria
Model	Group	Satorra-Bentler χ2 (df, P)	AIC	CFI		RMSEA (90% CI)

Nine scores (total MMSE included on “EF” factor (where 1 factor))

Higher-order model (one higher-order factor, two first-order factors). Consistent with EF as “higher-order” faculty	CERAD	3.32 (22df, )	40.68	1.0	0.039	0.00 (CI not computed)
	EPESE	8.69 (22df, )	35.31	1.0	0.040	0.00 (CI not computed)

First-order factors (no higher-order latent factor). Inconsistent with EF as “higher-order” faculty	CERAD	43.27 (24df, )	4.73	0.958	0.039	0.042 (0.021, 0.062)
	EPESE	32.96 (24, )	15.04	0.981	0.040	0.033 (0.00, 0.058 )

One-factor model: all test scores reflect a single factor	CERAD	141.91 (27df, )	87.91	0.748	0.078	0.097 (0.081, 0.112)
One-factor model: all test scores reflect a single factor	EPESE	97.48 (27df, )	43.48	0.848	0.064	0.086 (0.068, 0.105)

Ten scores (MMSE-WORLD on “praxis” factor, WORLD on “EF” factor (where 1 factor))

Higher-order model (one higher-order factor, two first-order factors). Consistent with EF as “higher-order” faculty	CERAD	18.91 (30df, )	41.09	1.0	0.052	0.00 (0.00, 0.007)
	EPESE	52.03 (30df, )	7.97	0.944	0.057	0.048 (0.024, 0.069)

First-order factors (no higher-order latent factor). Inconsistent with EF as “higher-order” faculty	CERAD	65.13 (32df, )	4.13	0.922	0.052	0.050 (0.033, 0.066)
	EPESE	63.93 (32, )	0.07	0.918	0.057	0.056 (0.35, 0.075)

One-factor model: all test scores reflect a single factor	CERAD	121.78 (35df, )	51.78	0.814	0.069	0.074 (0.060, 0.088)
One-factor model: all test scores reflect a single factor	EPESE	90.62 (35df, )	20.62	0.858	0.060	0.070 (0.052, 0.088)

Nine Scores (WORLD on “EF” factor, remainder of MMSE excluded (where 1 factor))

Higher-order model (one higher-order factor, two first-order factors). Consistent with EF as “higher-order” faculty	CERAD	22.73 (22df, )	21.27	0.998	0.035	0.009 (0.0, 0.040)
	EPESE	8.69 (22df, )	24.43	1.0	0.041	0.00 (0.0, 0.040)

First-order factors (no higher-order latent factor). Inconsistent with EF as “higher-order” faculty	CERAD	27.24 (24df, )	20.765	0.991	0.035	0.017 (0.0, 0.043)
	EPESE	29.60 (24, )	18.41	0.983	0.041	0.027 (0.00, 0.055)

One-factor model: all test scores reflect a single factor	CERAD	87.86 (27df, )	33.86	0.837	0.070	0.070 (0.054, 0.087)
One-factor model: all test scores reflect a single factor	EPESE	82.80 (27df, )	28.80	0.832	0.067	0.080 (0.060, 0.099)

fit indices have estimation procedures that are robust to distributional and assumptional violations except SRMR. The 90% CI for RMSEA in the higher-order model was not computable for either cohort.
All scores were from the baseline visit. In all models the latent variables derive their scale from standardization of their respective factor variances (set = 1.0).
Fit criteria: Satorra-Bentler χ2: general robust model fit statistic, with the associated -value for the degrees of freedom shown. Nonsignificant -value suggests “good” fit of model to data. AIC: robust Akaike’s Information Criterion; the lower, the better. CFI: Robust Comparative fit index; the closer to 1.0 the better; acceptable models have CFI .95. SRMR: standardized root mean square residuals, the smaller (and .09) the better. RMSEA: Robust root mean square error of approximation; the closer to zero (and positive) the better; acceptable models have an upper bound on the 90% CI .06.