Help with really f***ed up factorial ANOVA needed

[quaelgeist]

Hi

I really hope someone can help me with this.

The central analysis of the study i am working on is a factorial ANOVA (1 between subjects factor consisting of 4 different treatments and 1 within-subjects factor - a repeated measure)

My problem is with the statistical assumptions that have to be met so i can interpret the anova.

*) Normal Distripution is violated
*) Sphericity is violated
*) homogeneity of variances is violated


So...

Do certain conditions exist under which the factorial anova is robust against violations of all 3 of those assumptions? (propably not, i guess...)

Is there another method of analysis that can replace the factorial anova in this case? (maybe a "modern robust" one like the winsorized correlation for a normal correlation?)

If not, what analyses should i conduct to replace the violated factorial anova?


Thanks in advance for your help.

Michael.

[Masteras]

do not worry about the normality. Use the Greenhouse-Geiser or the other 3 statistics suitable when sphericity is violated. how many repeated measures do you have?

[quaelgeist]

Thanks for your reply

2 repeated measures - one pre-treatment and one post-treatment.

[Masteras]

My friend things are easy then. You do this Y=Xafter-Xbefore and then do a one way anova. In the anova if the constant can be assumed zero it will be the same as a t-test in y for y-bar to be zero.