Not getting the possibility to perform post-hoc testing for Bayesian ANOVA

[AugustOlsson]

Hi, I can't seem to perform post-hoc testing for Bayesian ANOVAs even though that function has (as I understood it) been implemented.

• JASP version: 0.11.1

• OS name and version: Windows 10 Education

• Analysis: Between Subjects BANOVA and Mixed BANOVA

• Bug description: When looking at the post-hoc test menue I do not get the possiblity to conduct post-hoc testing of the interactions, a function which has been implemented as I understood it. See screenshots

Best,
August

[vandenman]

Posthoc tests are possible for individual variables but not for interactions. Implementing it for interactions will require some additional work, in particular to get the multiplicity correction right. I do think it's a good idea though.

[AugustOlsson]

Hi,

Ah, I am sorry, then I somehow missunderstood which types of post-hoc tests that had been implemented. Thanks for remarking this as enhancement! It would be a great addition :). Thanks for all your great work with JASP:

[piolho-eefe]

Hi, is there any news on the implementation of post hoc tests for interactions in Bayesian ANOVA?
Thank you.

[JohnnyDoorn]

Hi @piolho-eefe ,

Unfortunately not, since we have had to focus on other issues for the next release. This will serve as a nice reminder though that it would be very nice feature! I will discuss with @vandenman about how best to tackle this for a future release.

Kind regards
Johnny

[tomtomme]

Related: #670
Also I assigned vandenman here too, since he was assigned to one of the duplicates I closed.

[tomtomme]

@AugustOlsson
Would you agree, that this one can be closed as a duplicate of #670
?

[AugustOlsson]

@AugustOlsson Would you agree, that this one can be closed as a duplicate of #670 ?

I guess it is almost the same thing? But I am not entirely sure as I am not to knowledgeble about simple main effects analyses. Dosn't a simple main effect analyses only look at the effect of each main effect in isolation (holding the other main effect constant) whereas a post-hoc test of an interaction would also (for example in a 2x2 faktorial design) compare if there is a difference in the dependent variable between lvl1 of independent variable 1 and lvl2 of independent variable 2?

In practice when implementing one of these the other one might though naturally follow?

Best,
August