Because group membership does not influence parameter estimates, any
concerns that generalising the survey findings to the population of waiting staff in
table service hotels may be adversely affected by survey non-response are
minimised.
As anticipated, because Model 5b is more complex (more observed variables and
more parameters to be estimated), the multi-group analyses for this model did
not produce such useful results. The findings for Model 5b are summarised in
Table 7 - 39 where it can be seen that only the ‘age’ grouping produced a fully
satisfactory set of multi-group models based on the estimates for model fit.
The finding for the age grouping leads to the conclusion that age does not
moderate the measurement or structural relations found for Model 5b.
Grouping CFA model
fit^2
CFA
invarianceSEM model
fit^2Structural
path
invariance
Gender χ^2 sig^ Partial metric^ χ^2 sig^ Invariant
Age Good fit^ Full metric^ Good fit^ Invariant
Length of
employment χ
(^2) sig Full metric χ (^2) sig (^) Invariant
Part- /Full-
time χ
(^2) sig Full metric χ (^2) sig (^) Invariant
Respondent
origin χ
(^2) sig Partial metric χ (^2) sig (^) Invariant
Table 7-39 Moderator analysis for Model 5b
For the remaining four groups, as with the multi group model based on gender in
Model 2, although the p value for the χ^2 estimate was significant (< 0.05
indicating less than satisfactory model fit) the CFI values were all satisfactory at
0.96 and all RMSEA estimates were satisfactory at below 0.06.
Based on these findings, it is concluded that age does not moderate any
relationships between constructs in Model 5b. For the other four multi-group
models (gender, length of employment, part-time/full-time and respondent
origin) the findings suggest that group membership does not moderate effect
sizes between constructs in Model 5b. However, although the CFI and RMSEA
estimates for model fit were within acceptable ranges, the significant χ^2 p value
estimates for model fit for each of these four multi-group models prevent firm