MANOVA Post Hoc - why not pairwise comparisons?


New Member

When running ANOVA post hoc, we compare all the pairs of groups using t-test=ANOVA with two groups and corrected alpha or Tukey HSD.

When running MANOVA post Hoc, I didn't see the standard procedure is to compare all the pairs of groups using MANOVA with two groups
with the same statistics used for MANOVA (Hoteling's trace, Pillai's trace, etc)

I saw the following:
1. Several ANOVA tests per DV - it doesn't make sense as the primary method, since we lose the advantage of MANOVA, if ANOVA over specific DV is not significant, it doesn't say that the DV doesn't contribute to the MANOVA due to covariance. (helps only to say that DV is significant.

2. Contstrats - makes send like in ANOVA, but specific for one check. (of course, you can perform manny contrasts)

3. Linear discriminant analysis (LDT) - why is it better?