Does frequency distribution matter when using nominal/ordinal data?

#1
I want to do relationship analysis between different variables from questionnaire results that I've just collected and I've recently learned that you need to make sure certain assumptions are fulfilled before using certain analysis tools such as needing to have a normal distribution for your data. In my case I want to study the relationships between different question responses, I plan to use chi-square test for independence or ordinal regression for my analysis. But does the frequency distribution of the responses matter if most of my results are nominal/ordinal and I'm using a stats tool such as chi-square test for independence?
As you can see from my below results, one question has one kind of distribution, the other another kind. If frequency distribution doesn't matter for these types of variable data, when does it matter? Thanks!

 

hlsmith

Omega Contributor
#2
To my knowledge there is no such assumption. There is the requirement, chi-sq, that each cell has an expected frequency of 5 or greater. Also not sure if some would frown on having the "don't know" group included, especially in ordinal regression where it inclusion seems inappropriate.