The guidelines for moderation testing provided by Hair et al. (2006: 870- 876 )
were followed to assess what (if any) differences on effect sizes were observed
when respondents were grouped by gender, age, origin, full- / part-time,
temporary/permanent tenure and length of service.
The procedure for moderation testing follows a similar procedure to that
described for the multi-group split-sample cross-validation of the model in Section
7.12. To test for moderating effects in a structural model, it is first necessary to
establish either full or partial metric invariance in the measurement model. This
involves checking that internal factor structures (how indicator variables load onto
their respective factor) are consistent across groups (Hair et al. 2006: 873). It is
possible to move on to testing the structural model for invariance in path
coefficients where the associated multi-group measurement model demonstrates
adequate fit (based on the thresholds for model fit described earlier in Figure 7 - 3 )
and full metric invariance (all indicators equivalent across groups) or partial
metric invariance (at least two factor loading estimates equal across groups) (Hair
et al. 2006: 825). Chi square difference tests are used to assess invariance (or
its opposite, non-equivalence) at both measurement model and structural model
stages in the same manner as described in Section 7.12.
Unlike the split sample analysis for model validation undertaken in Section 7.12,
where group sizes were almost equal (106 cases versus 107 cases), when the
groups are defined based on the demographic variables, the groups sizes are not
always well-balanced. Also, owing to missing values in the demographic data,
total sample sizes are lower. Table 7 - 37 summarises the group sizes and extent
of missing data for the demographic variables.
Variable Demographic categories
Gender
(missing n = 4)
Female Male
116 93
Age
(Missing n = 4)
Less than 24 years 24 or more years
114 95
Length of employment
(Missing n = 5)
Less than 1 year More than 1 year
97 111
Part- /Full-time
(Missing n = 12)
Part-time Full time
78 123
Respondent origin
(Missing n = 2)
UK Non-UK
76 135
Tenure
(Missing n = 30)
Seasonal/Temporary Permanent
31 152