That is, the model of this study includes the five latent variables with 12 observed
variables. The variables loaded neatly onto five factors, as predicted. The following
model is the estimation of the author's model. The pattern of fixed and free parameters in
a structural equation model defines two components of the general structural equation
model: the measurement model and the structural model. The measurement model is that
component of the general model in which latent variables are prescribed (Hoyle, 1995).
The results of goodness of fit on the measurement model are shown in Table 4-9. The fit
of the original, hypothesized model (Figure 3-1) was adequate because the chi-square
was statistically significant (X2 (66) = 1708.237), and the additional fit indices showed
that the model has fairly good fit (NFI= .937, NNFI= .943, and RMSEA= .079).
However, as discussed in the previous section, there is a fatal problem in this proposed
measurement model. There are very high correlations between the two variables in rapid-
naming and vertical/horizontal eye movement variables. For example, DEM vertical test
scores are highly correlated with digit rapid naming and letter rapid naming (r= .849 and
r= .805). This is an evidence of multicollinearity in this model. Therefore, new
constructs were re-identified. Table 4-9 presents a modified constructs and goodness of
fit results for the newly identified measurement model. As predicted, a measurement
model fit became much greater with this modification (X2 (45) = 78.160, NFI= .954,
NNFI= .970, CFI= .980, SRMR= .050, RMSEA= .057).
Therefore, within the measurement model, all of the measured variables were
significant indicators of their respective latent variables, and no modifications of the
measurement model were needed since all of the fit indices met their criteria for good fit.