Page 12 - IJPS-8-1

International Journal of
Population Studies Validity and reliability of Mini-Mental State Examination in older Chinese

the seven-factor was better than the six-factor solution, the coefficient of 0.64, indicating that an older adult that has
orthogonal seven-factor was included to examine whether a high recall was also high in language capability. These
an unrelated of the seven-factor constructs of MMSE was results supported the proposed theoretical CFA that
better than the seven-factor oblique model. While the one- these seven MMSE domains were distinct yet associated.
factor CFA model was that all the 23 items were grouped The Zumbo’s alpha coefficients for all the six MMSE
under one overall MMSE construct, a second-order seven- constructs were all high in reliability with values all
factor CFA also indicated an overall MMSE construct higher than 0.89.
however loaded under the 7 MSSE constructs instead of Figure 1 shows the graphical representation of the
the 23 MMSE items. seven-factor oblique CFA model. These seven factors

Table 2 shows the results of the proposed seven-factor were orientation, short recall, delayed recall, calculation,
oblique CFA model and the four competing CFA models. language, comprehend instruction, and visuospatial. The
The fit indices CFI and TLI of the hypothesized seven- standardized factor loading coefficients were indicated
factor oblique model (1.0; 1.0) were higher than the four on top of the arrow that ran from the construct to the
competing models. Similarly, the RMSEA and SRMR of MMSE items. All the factor loadings were high in value
the hypothesized model also indicated a better fit than except for visuospatial. These high loadings indicated
the four competing models with the lowest values (0.009; the high association of the MMSE items to the respective
0.033). These results indicated the proposed seven-factor latent factor MMSE construct. The error residuals were
oblique model fitted better than the four competing printed after the items, on top of the arrow that ran from
models. Table 3 displays the latent correlations of the error terms (E1 to E23) to the MMSE items. These error
seven factors together with their reliability indicator, residuals were low in value also indicating that the MMSE
Zumbo’s alpha. Since all the MMSE items were binary items were low in measurement errors when they were
coded, the ordinal Zumbo’s alpha was reported (Zumbo, loaded into the appropriate MMSE construct. The double
Gadermann, and Zeisser, 2007). The factor correlations arrows that ran between the seven constructs indicated it
for all the seven domains of MMSE were all positive. represents an oblique CFA model.
These positive coefficients indicated an older adult that
possessed a high MMSE construct in one cognitive 3.4. Factorial invariance
function domain tended to also possess high in another Table 4 summarizes the results of factorial invariance of
domain. For instance, the short recall was moderately the two age groups. The CFI for all the four invariance
correlated with language with a positive correlation conditions was all at a high value of 0.999. Similar results

Table 2. Fit indices of hypothesized cfa and competing models.

Model χ 2 df CFI TLI RMSEA SRMR
Hypothesized seven-factor oblique 341* 209 1.000 1.000 0.009 0.034
Competing seven-factor second order 605* 223 1.000 0.999 0.014 0.044
Competing six-factor oblique 609* 215 1.000 0.999 0.015 0.050
Competing one-factor 9196* 230 0.976 0.973 0.069 0.123
Competing seven-factor orthogonal 77470* 230 0.789 0.768 0.201 0.467
χ , Chi-square statistics; df, degrees of freedom; CFI, comparative fit index, TLI, Tucker-Lewis index; RMSEA, root mean square error of approximation;
2
SRMR, standardized root mean square. *P < 0.01.
Table 3. Factor correlations and Zumbo’s alpha for the seven domains of MMSE.

MMSE construct 1. 2. 3. 4. 5. 6. 7. Zumbo’s alpha
1. Orientation 0.93
2. Short recall 0.53 0.96
3. Delay recall 0.48 0.57 0.94
4. Calculation 0.57 0.58 0.50 0.98
5. Language 0.56 0.64 0.54 0.63 0.91
6. Instruction 0.32 0.40 0.34 0.40 0.51 0.89
7. Visuospatial 0.38 0.37 0.35 0.50 0.39 0.29 -

Volume 8 Issue 1 (2022) 6 https://doi.org/10.36922/ijps.v8i1.1285