I am looking for advice on what to do when percent error data (or just error rate) violates the assumption of normality for a 2x2 repeated measures ANOVA (N=51). I have already removed outliers, and that does not "help" the normality issue. I have also check the square-root and the sin transformations, both of which still show violations for most of the unique conditions.
I'm using the shaprio test and the qq-plots. The shaprio test shows p < .05 for most of the conditions, but the qq-plots really only show one unique condition that might be driving the violation -- the others mostly follow the line.
I know that you can look at non-parametric test when one-way anovas violate this assumption but there doesn't appear to be a non-parametric method for a 2x2 repeated measures anova design.
If you have any suggestions on how to address this, please let me know. I am aware that Anovas, in general, are robust to this violation but if you look up the literature on this there seems to be two camps: those who say it doesn't matter a lot and those who say it matters. Anyways, looking for some statistical advice/suggestions/etc.
Thank you for your time and consideration.
