Dear mailing list,
My question in short is: When modelling within-subject factors (WS-ANOVA/flex factorial), why is equal variance assumed for the "subject" factor and why does changing this setting in flexible factorials affect the results dramatically?
The last mention of this question seems to be an older poster from 2012 (https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=spm;1bdb4cfe.1210) yet no conclusion was reached here, in fact, more questions are raised. The assumption in this post (and now in mine) is that if the variance is truly equal, then "applying the correction" shouldn't affect the results much; yet Z/T values are inflated drastically.
Any clarification would be greatly appreciated.
-Rik
|