I started working on a reprex of this but I'm honestly not sure if it would help yet.
Essentially--I'm working on a personal project trying to familiarize myself with RStudio (I'm a beginner) and multivariate statistics. I found a dataset that has a bunch of Likert scale measures and for each measure, there is a mean score for each participant. So each person has a mean Likert score for each measure. It is a pre and post test but since I'm just trying to learn, I decided to treat the groups as a 'no treatment' and 'treatment'.
My problem(s) are that this set is failing all assumptions tests for MANOVA. I tried log transformation and that helped get "better" results (as in, it passes some assumption tests) but it's my understanding that log transforming Likert scale data is inappropriate. Is it also inappropriate for a mean Likert score of a whole measure?
There are other multivariate analyses to do but I figured I should try MANOVA as it's the simplest and I happen to have a tutorial for it...
tests failed:
Shapiro-Wilks (though the Q-Q plots of the groups looks okay-ish to me)
boxM
Any feedback or thoughts would be so welcome. If that includes making an example, I can try that as well.
Thanks