So... Can you please tell me which approach is valid?
I have a set of 3 repeated measurements for 2 groups, I want to test for differences between measurements and for differences between groups
First measurement is normally distributed, the others are not...
What is the right way to test for differences between measurements and differences between groups?
a) do the Friedman test for each group, and then test each measurement between groups (first Student's t, and then Mann Whitney)
b) do the RM-ANOVA with existing data
c) do the Box-Cox transform and then the RM-ANOVA
Thanks
Repeated measures, non-normal distribution
Re: Repeated measures, non-normal distribution
hi,
i think it would depend on the sort of research question you have. you have a 3x2 mixed design, but often one comparison is more important than the other - so that's worth considering.
you might transform the data, but i'm of the view that you need a principled reason to try a transform (rather than just trial-and-error-ing it).
so i think what the best choice here will probably depend on a number of other things. i imagine all three of these approaches could be appropriate depending on the circumstances.
cheers
jonathon
i think it would depend on the sort of research question you have. you have a 3x2 mixed design, but often one comparison is more important than the other - so that's worth considering.
you might transform the data, but i'm of the view that you need a principled reason to try a transform (rather than just trial-and-error-ing it).
so i think what the best choice here will probably depend on a number of other things. i imagine all three of these approaches could be appropriate depending on the circumstances.
cheers
jonathon