If 'test that all slopes are zero' is non-significant, do levels of IVs need to be checked? I've run a univariate ordinal regression in Minitab and my 'test that all scopes are zero' value is non-significant at my chosen alpha (p=0.141). According to the Minitab guide, this value represents "whether at least one of the predictors in the model has a statistically significant association with the response events". This would indicate to me that the single variable in the model is non-significant.
When I look at the different levels of the IV though (occupational categories in this case), I can see significant differences between some occupations with p<0.05.
I'd like to confirm that the following interpretation of this situation is correct:

*

*The IV (occupation) has no significant influence on the DV per the test of slopes.

*Despite the significant differences between some levels of the IV, there is no need to check each level against every other level, since the IV itself is non-significant.

*Only if the test of slopes was significant would it become necessary to check every IV level against every other level.

Thank you!
 A: *

*It's possible that your study was simply too small to detect a significant difference among occupations. That would happen if you had a lot of occupation categories versus the total number of cases. You can't really say "occupation has no significant influence on the DV," just that your study didn't demonstrate an association of occupation and the DV.


*Standard procedure would be to stop once you didn't find a "significant" overall test for slopes. It wouldn't be appropriate to claim that any differences that you found among occupations are significant. But if you are interested in pursuing this field of study further, you might want to examine those differences for yourself as a guide to design of future studies.


*It's never necessary to do all pairwise comparisons even with a significant overall slope result. The more comparisons that you make, the more stringent corrections for multiple comparisons that you need. It's typically best to have a limited number of specific comparisons that you planned to make before looking at the results.
One suggestion: it's possible that you could find something of interest if you cut down on the number of occupational categories you use. Look at the current categories (without looking at their relationships with the DV) and see if it would make sense to combine some of the current categories based on what you know about the occupations. For example, if you have "bakers" and "cooks" as separate categories, maybe combine them into "restaurant kitchen workers."
