Post-hoc testing for interaction in 2way BANOVA + Effect sizes
Hi,
Two questions below I've been pondering lately.
1.) Is there any way to do a post-hoc tests for interactions in a between-subjects 2-way BANOVA in Jasp . From what I can tell I can only post-hoc testing the main effects. I could of course look at if the credible intervals for each group overlap with the mean of another, but that would not correct for multiple testing and might thus be problematic. Or is it feasible?
2.) How do you approach effect size within the Bayesian ANOVA framework. To my understanding JASP does not provide any direct effect size (such as partial eta squared for frequentist ANOVA) which describes e.g. amount of varriance explained by main effect(s) and interaction effects. Generally I guess that the BF factor is correlated with the size of the effect, but to my understanding the BF-factor only provides evidence for an effect being present, and not for the size of the effect. How would you describe the effect size, and is there any neat function in JASP which could be used that I havn't found?
Best,
August
Comments
I realised that this might be a bug, because when reading the notes I saw that post-hoc testing had been implemented, at least for Mixed-BANOVA - and I don't get the possibility to do that either. I've opened a bug-report on the matter here: https://github.com/jasp-stats/jasp-issues/issues/588
Applogies if I am missing something here..
Best,
August
Dear August,
Thanks for posting the issue and sorry for the tardy reply. About the effect size question: I do think we have recently added posterior distributions in JASP, but I am not sure they are for traditional effect size measures. I'll ask the team.
Cheers,
E.J.
Dear EJ,
No worries. Turns out that I misread the notes and post-hoc testing had not been included for interactions. So my post on the matter has been moved to enhancement. It would be a great inclusion though.
Best,
August