My apologies for posting so quickly yet another question, but hopefully you can help me out once more. I'm still analyzing my survey data that was held to find out what students though of a course. To check for differences I'm comparing the questions based on location (this is what my firs questions was about here), gender and how good the students find themselves in a certain area.

For the gender comparison I'm using a Chi-square test for items on nominal level, an independent Mann-Whitney U test for items on ordinal level (5 point scale), and an independent samples t-test for items on an interval/ratio level.

Q1: Is this actually correct? I think so, but might be good to hear from others.

My bigger problem comes with the comparison based on how good the students ranked themselves. This was a 5-point scale question: To which extend to you agree with the statement: I am good in number subjects. Fully disagree O O O O O Fully agree.

To compare the results on this item with other ordinal items, I'm not sure to use the Wilcoxon Signed Rank test or a correlation coefficient of some sort. The same doubt exists when comparing this with other scale variables, should I use a Kruskal Wallis test or a correlation coefficient. I'm guessing it's actually a correlation coefficient, since this would make it possible to say things like "Students who perceive their own number skills higher tend to evaluate the available material lower.".

Q2a: Should I use a rank test (Kruskal Wallis and Wilcoxon Signed), or a correlation test and coefficient?

Q2b: In case of a correlation, I think it should be a Spearman's rho, is this correct?

Apologies for the many questions, but I'm starting to get a bit lost in all the possible options.

Thanks in advance for your response.

Kind regards,

Peter

p.s. any help still with my previous posted question is of course also still welcome.