Comparing nominal and ordinal variables .

Question

Hi,

I am conduction a prospective study, in which I follow up patients using a questionnaire, and I repeat the same questionnaire in different stages of the follow up.

The collected data from the questionnaire are nominal and ordinal variables.

I want to know what kind of progression my patients had during the follow up by comparing their answers o the same questionnaire in the different stages of follow up.

My question is: what test to sue to test my result for statistical significance?

Answer 1

Hello,

I love this question.

When working with ordinal and nominal data you have some options when it comes to statistical significance. Remember, statistical significance asserts that the sample's (or observed cohort in your case) distribution is different from the original to the extent that the results are unlikely to be due to chance.

One option is to treat questionnaire items as groups and calculate the sum or average. Then, compare those means in an independent T-test. You could use Factor Analysis techniques with the ordinal data to examine the variance. However, nominal data becomes a bit more tricky to group. I have see researchers group nominal data based on some theoretical argument. But, this is not as strong in some fields.

All this to say, once you make some decisions you could run a paired T-test or McNemar's exact test. The former would treat grouped items as continuous and the latter would compare the nominal/ordinal data.

I hope this helps!