ANOVAs v's Planned comparison ANOVA

#1
Hello again,

I'm wondering if it is ok to carry out an omnibus anova and use post-hoc tests if you have some idea of where you expect to find differences?

Is there a big difference between planned comparison anovas and typical one way anovas.

how do you get eta squared if you carry out planned comparisons? I just usually get it from the omnibus anova...

I have 3 groups AD, DS & ID-P and my hypothesis is that there will be differences between both DS and ID-P and AD on all measures and I don;t want to combine groups DS & ID-P. So I want to see if there are differences between the three across a number of measures and then between all 3 groups. Up to now I have been doing ANOVAs and looking at Tukey HSD 'post-hoc' test eventhough I have some idea before I started that there would be differences in certain groups. Is it incorrect to use post-hoc tests then..

I keep getting conflicting advice which is confusing...

Would appreciate any advice...

Thanks
M
 

JohnM

TS Contributor
#2
It's not "incorrect" to use ANOVAs and post-hocs when you have apriori theories/hypotheses, but you're putting yourself at a disadvantage because Tukey post-hocs are not as powerful as plain-old t-tests.

If you have constructed specific comparisons in your theory, prior to collecting data, then don't do an ANOVA followed by post-hocs - just do the planned comparisons with the test that is most powerful.

ANOVAs and post-hocs are only necessary when you're not really sure where you'll find differences.