the nested modeling test, thereby assessing the need for different paths (Anderson and
Gerbing 1988). After incorporating the modifications based on the information provided
by the Lagrange Multiplier (LM) and Wald (W) statistics, which assesses the effect of
freeing a set of parameters simultaneously, structural model was analyzed.
The Lagrange Multiplier (LM) statistic allows one to evaluate the effect of freeing
a set of fixed parameters in a subsequent model [referred to by Bentler (1986) as a
forward search] to improve the fit indices significantly. The Wald (W) statistic was used
to evaluate whether the free parameters in a model are necessary in a statistical sense. It
indicates which parameters in a model should be dropped [referred to by Bentler (1986)
as a backward search] in order to improve the fit indices.
As suggested by the Lagrange Multiplier (LM) test the error of four items (i.e.,
SAT 1 and SAT 2; SAT 1 and SAT 6) of the satisfaction construct was allowed to covary.
The results indicated a significant improvement in fit indices in the final revised
structural model compared to the proposed/base model. Table 4.8 shows the model fit
indices of the final revised structural and the proposecLbase model after modifications.
Multiple fit indices, as suggested by Hu, Bentler, and Kano (1992) and
Schumacker and Lomax (2004), were used for assessing the model fit of the final revised
structural model. The results indicated that the overall fit indices of the structural model
(Figure 1) were excellent, indicating a close fit. The Satorra Bentler Chi square ratio
(S-B %2 /df) statistic of 2.23 was found to be within the acceptable range of 3 (Hayduk
1987). The other fit indices also indicated a close model fit. All the fit indices, including
incremental fit index (IFI), comparative fit index (CFI), and non-normed fit index (NNFI)
values, were above .90 as recommended by the literature. The root mean square error of