Bayes ANOVA - Any chance to adjust prior probabilities?
Dear E.J.,
finally I analysed my own data for a study, that I'm supposed to replicate. Remeber the orignal effect was really huge (p< .001, ηp^2 = .136; BF10 =800000...)?
Now, in the replication data, the ANOVA shows no evidence for anything: p = .075, np^2 =.057; BF01=1.66 with a power of 60% (N=122).
Now, what would be a sensible answer to the question, whether the replication was successful? On the first glance, the previously set criteria (significant p value; CI of Eta to include original effect size) were not met. But what if I took into account the huge effect that showed in the original study and altered the prior probability of H1 accordingly? Actually, everybody expected the effect to replicate with a 90% probability. Is there any way to do that with JASP?
Thanks so much,
Jan
Comments
Hi Flaihai,
If you are interested in accounting for knowledge gained in the first study, you can use a replication Bayes factor (for a simple application you can read DOI:10.3758/s13428-018-1092-x), assuming this is a direct / exact replication.
Good luck!