The structural model is found to fit well based on all fit measures except for chi
square, which has a p value of 0.011. Examination of the standardised residual
covariance matrix (SRCM) reveals that this lack of good fit can be accounted for
by the large SRC values shared between WV6 and JP1. There is no item content
overlap between these indicators and an examination of the standardised residual
covariance matrix reveals that WV6 has a greater amount of unmeasured
variance associated with it (30) than does JP1 (at 23.3). The removal of WV6
(Work in general is interesting and satisfying to me) does not significantly reduce
the substantive content of the Work Value construct.
The model is re-estimated as SEM 4:2: the structural parameter estimates and
model fit statistics for SEM 4:2 are illustrated in Figure 7 - 18.
Figure 7-18 Estimates for Model 4 (SEM 4:2)
A further measurement model (CFA 4:8) is estimated without the WV6 indicator
to ensure that model fit and convergent and discriminant validity are upheld.
Model fit is satisfactory (χ^2 = 93.572, d.f. = 94, sig = 0.493; RMSEA = 0.0000
(0.036; 0.000; pclose = 0.998); CFI = 1.000; SRMR = 0.0386; and Hoelter’s CN
= 267). Convergent validity is maintained (all AVEs >0.5 and all CRs >0.07) and
the highest squared correlation estimate is ML EPA at 0.325 and the lowest
AVE is 0.655 (for JP).
The assessment of distributional normality indicates that the data demonstrate
multivariate non-normality (multivariate kurtosis critical ratio = 16.6). Both the
CFA 4:8 and SEM 4:2 models were, therefore, re-estimated using the bootstrap