[Q] How to correctly test assumption for equal variances for a two-way model? (Can't add a flair)
I'm confused about what to do for my data. I can't use the common equal variance tests in my software because those are for one-way models only. May I ask for suggestions?
In JMP, I've discovered a heteroscedacity test (no name for the method, just heteroscedacity test) where it shows the variances of the two one-way models and the whole two-way model graphically with no p-value involved. However, some results suggest that variances are not equal in some one-way models, but when you look at the whole model, variances are equal. If this method is correct, should I just use the whole model graph when testing for the assumption for equal variances? Or should all the one-way models and two-way model show equal variances for the data to achieve the equal variance assumption?
Residuals are used in testing for assumptions, not raw data, right?
Suggestions are highly appreciated.