Figure 7-14 Parameter estimates and model fit for SEM 3b
Checking the AMOS output for multivariate normality finds that the CR value for
multivariate kurtosis of 14.9 is indicative of multivariate non-normality and,
accordingly, the model is re-estimated using the bootstrap procedure. The
bootstrapped estimates and their associated confidence intervals indicate that
these parameter estimates are robust under the conditions of multivariate non-
normality and the model can be accepted.
A final task remains for Model 3b. Specifically, it is necessary to test the
mediation effect of the new Employee Positive Attitudes (EPA) construct.
Testing the mediator effect
Testing to evaluate the full or partial nature of the mediator effect of (Employee
Positive Attitudes) EPA on the ML→JP relationship was undertaken following the
same procedure from Hair et al. (2006: 866-8701) that was employed for Model 2
above (Section 7.4).
Firstly, a constrained model with no direct effect (ML→JP) was estimated (SEM
3b 1 ). This constrained model produced the following fit statistics: χ^2 = 69.552,
d.f. = 62, sig = 0.238; RMSEA = 0.024, CFI = 0.995, SRMR = 0.0700. The chi
square difference (Δχ^2 ) between models SEM 3b (χ^2 = 60.084) and SEM 3b 1 (χ^2 =
69.552) was calculated at 8.748 (69.552 – 60.804). With 1 degree of freedom
this yields a p value of 0.0031. Accordingly, (because p <0.01) full mediation of
the ML→JP path by EPA is not supported.