0
$\begingroup$

I'm running a repeated measures ANOVA with a 3-level DV. Now what I've read is that you only look at the post-hoc test if the within-subject (overall) effect is significant. What about the within-subject contrast: what if the effect is not significant, but certain levels in the contrasts are?

Two (related) questions: 1) would you report the significant contrasts despite the not significant effect? Or is it just as with post-hoc that you would ignore the contrasts completely if the effect is already not significant. 2) can you look at the post-hoc test if certain contrasts are significant?

$\endgroup$
0
$\begingroup$

Recall, in repeated measures ANOVA, there is a within-group effect $(W)$, a between group-effect $(B)$, and the interaction $(B*W)$. The within-group effect reveals how the response values change over time within each group. If responses never change within a particular group and are constant, then $W$ won't be significant. For the between-group effect $B$, this basically compares ovarall mean response differences between groups. The real focus is the $B*W$ interaction, since it lets you know if the groups differ in their reponse over time, such as linear, quadratic, or U-shaped dose-response curves. For contrasts, they are by definition, all post hoc, so you can do whatever you want. Yes, you can report significant all-possible mean comparisons between groups and times whatever fashion you want -- just make sure you state that e.g. "during post-hoc analyses, we identified significantly different means between sub-groups x,y,z and/or times 1,2,3." The key is to preface your statement with "post-hoc," after which you can literally perform any tests you desire.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.