Using the groups defined by the demographic variables outlined in Table 7 - 37 ,
Models 2 and 5b were examined for between-group invariance. Considering that
the smaller groups for part-time/full-time (n = 78) and respondent origin (n =
76) fell some way below the minimum recommended sample size (100) for SEM,
these two groups were not expected to perform well in terms of producing multi-
group models for the more complex Model 5b.
It was hoped, however, that for the simpler Model 2, adequate model fit might be
achieved. Multi-group analyses based on the tenure groupings
(seasonal/temporary n = 31 versus permanent n = 152) were not attempted on
either Model 2 or 5b since 31 cases is below any recommended minimum for case
numbers in SEM analysis (see Section 6.5.2).
The results of the multi-groups analysis for Model 2b are summarised in Table
7 - 38. The only poorly-fitting multi-group model found was that based on the
gender grouping. For this model, although the χ^2 estimates were significant (in
the measurement model, p = 0.039 for the unconstrained model and 0.021 for
the constrained; in the structural model 0.014 unconstrained and 0.011
constrained), the CFI and RMSEA estimates were adequate (CFI >0.96 and
RMSEA <0.06). For the gender-based multi-group model, because the χ^2 p value
was < 0.05 (i.e. not satisfactory) but the CFI and RMSEA values are satisfactory,
the findings of invariance measurement and structural models can be cautiously
accepted. For the remaining four multi-group models, model fit was satisfactory
according to all fit measures and none of the groupings moderated the effect sizes
for structural paths in Model 2.
Grouping CFA model
fit^2
CFA
invariance
SEM model
fit^2
Structural
path
invariance
Gender χ^2 sig Full metric χ^2 sig Invariant
Age Good fit Full metric^ Good fit Invariant
Length of
employment Good fit^ Full metric^ Good fit^ Invariant^
Part- /Full-
time Good fit^ Full metric^ Good fit^ Invariant^
Respondent
origin Good fit^ Full metric^ Good fit^ Invariant^
Table 7-38 Moderator analysis for Model
