# What statistical test should I use for subgroup analysis/correlation?

I am a conducting a four group randomized survey-based study.

In the first group, I explain a situation, and then ask respondents their willingness from 1-10 to take a medication.
In the second group, I explain a situation, and then ask the respondents their willingness from 1-10 to take a medication.

The same repeats for group three and four.

I believe if I wanted to compare the "willingness responses" between each groups, I could use a basic Kruskal–Wallis test.

However, for each respondent, I will also be asking them: 1) their age; 2) gender; 3) educational level; and 4) how long they have suffered with their disease.
I was wondering, what is the best test to use to compare if patients' subgroup categorization has an effect on their willingness to take a medication?

I took a closer look and did some digging, what do y'all think of ordinal logistic regression?

• Are the 'explanations' different for the four groups? Are these what you really want to test? – David Smith Jun 5 '17 at 14:22
• Logistic regression is usually used for binary outcomes, not a scale of 1 to 10. There are extensions to ordered categories, but there are usually a small number, say, up to about 5. – David Smith Jun 5 '17 at 14:23
• Yes, the explanations are different in the four groups. The wording on how the situation is presented is different. We then subsequently ask the individual to rate his willingness to take the medication from a scale from 1-10.Yes, this is what I really would like to test. – TheFermat Jun 6 '17 at 11:26
• What range of responses were actually used by the subjects. Did you actually get ratings from 1 to 10 or did they actually lie in a more restricted range such as 7 to 10? – David Smith Jun 6 '17 at 14:09
• Their responses were non uniform, some with 1-4 and others with 7-10. Thanks so much for your help so far! – TheFermat Jun 7 '17 at 15:41