Calculating Delta for Post Hoc Test

AgnesB

Hello!

Does anyone know if there is a way to calculate Delta for a post hoc test following a Baysian ANCOVA?

Thanks! ☺️

EJ

The effect size estimate under H1 would remain the same, as would the Bayes factor. The post-hoc character of the test affects only the prior model odds. Now if you wanted to obtain a model-averaged estimate of effect size using both H0 and H1, then that estimate would change (and be shrunk more to 0, because H0 now has a bigger weight). See also https://psyarxiv.com/h6pr8/

Cheers,

E.J.

AgnesB

Thanks, that ref is really interesting perhaps a little bit over my head haha!

so you know when you run a Bayesian independent samples t-test in JASP you get a delta value, what I was hoping for is a way to achieve the same thing, but in the context of a post hoc pairwise comparison following a Bayesian ANCOVA. Is this possible?

Thanks so much!

Agnes

EJ

Hmm I am not 100% sure what you would like. I assume that with "delta value" you mean the posterior point estimate of effect size? If so, see my earlier response. If not, could you clarify?

Cheers,

E.J.