permanent / seasonal / temporary;
length of service; and
respondent origin (UK / non-UK).
The analysis was carried out using the SPSS Missing Value Analysis (MVA) module
and revealed some minor fluctuations in the proportions of missing data
attributable to age (and its effect on non-response to JS5) and length of service
(and its effect on non-response to EM3). However, the absolute volumes of
missing data in these variables involved are so small (the JS5 variable is missing
5 out of 213 total observations and EM3 is missing is missing 3 out of 213) that
these fluctuations cannot be regarded as meaningful in a statistical sense. To
investigate this further χ^2 (chi square) tests were carried out on age/JS5 and
length of service/EM3. These tests confirmed that it is not possible to infer
statistical meaningfulness on these findings as the assumptions of the χ^2 test are
not met (i.e. the proportions of expected counts less than 5 are >20 per cent).
Using the SPSS MVA module once again, the item statement data were tested for
their MCAR (Missing Completely at Random) status using Little’s (1988) MCAR
test. This procedure tests the hypothesis that the missing data are missing
completely at random (MCAR) and a p value of <0.05 means that this hypothesis
should be rejected – meaning that the missing data cannot be classified as MCAR.
The analysis produced a p value of 0.002 and so the hypothesis that the missing
data are MCAR is rejected.
In conclusion, the data meet the assumptions for MAR but fail the test for MCAR.
Additionally, the characteristics of the missing data satisfy the four rules of thumb
described by Hair et al. (2006: 55-56).
6.6.2 Distributional normality
As noted in Section 6.3 above, none of the measured variables violate Curran et
al.’s (1996: 26) bounds of 2 for univariate skewness or 7 for univariate kurtosis
(see Appendix II for details).
Multivariate normality is measured using AMOS’s critical ratio (C.R) for
multivariate kurtosis which is equivalent to Mardia’s (1970) normalised estimate
of multivariate kurtosis (Byrne 2010: 104). C.R. values greater than 5 can be