more strongly onto the Verbal Aggression factor (.33 ≤ loading ≤ .64)
than Hostility (.01 ≤ loading ≤ .17). Another Hostility item loaded .30
on its hypothesised factor, with a substantive loading on all other factors
(.18 ≤ loading ≤ .28). Another Hostility item loaded only .14 on Hostility
but .44 on Anger. This suggests particular problems with the Hostility
construct in this sample. There were also a number of issues with two
Verbal Aggression items and one Physical Aggression item substantively
cross-loading.
In order for gender difference results to be considered valid and
reliable, gender invariance must be assured. Configural invariance was
supported in both AQ short forms, indicating that males and females
held the same basic perception of trait aggression, distinguished
between the four factors, and identified the same items which loaded
onto these four factors. Metric invariance was also supported, indicating
that males and females used comparable conceptual frames of reference