A chi-square test on this data gives me X2 (11.1,3, p = .011). (or if
done on %s X2 (8.78,3, p=.032)
Now I would like to describe the %s and I was wondering if there was a
way of saying these, in turn, are significant.... for example,
"Taxonomy 7 errors were accountable for 44.2% of all rotary wing
incidents, which was approximately 2x higher that taxonomy 7 errors
incurred in fixed wing incidents (23.3%). A further 'test of
association' (?) showed these differences were indeed significant (p=.
However you could try fitting a logit link model to your data. In systat at least the output compares each level of the DV to each other giving it an odds ratio. You can then take the differences between the odds ratios of the groups you are interested in.