ANCOVA in RCT
Hello JASPians =)
The recommended frequentist analysis for a RCT with parallel groups and two measurements (Pre-Post) is an ANCOVA with Baseline (Pre) as covariate and the outcome (Post-Score - or change score which is equivalent in a frequentist's approach). I want to do a straightforward comparison of the bayesian equivalent.
My idea: Add the Pre-term to the null model.
Would that be the right approach? In that case the p-value would be .034 and BF = 2.01 with samples of 12 and 13 per group...
Comments
Hi Rob,
Yes I think that makes sense. To illustrate, I did the 4 analyses in JASP: https://osf.io/36mxs/
It contains the frequentist anova with post-test and pre-test as covariate, compared to the ANOVA with the change scores. Then I repeated it with the Bayesian ANOVA, and they all give very similar results.
Kind regards
Johnny
Hey Johnny,
thank you very much for your feedback! In the bayesian analysis the DV is pre and the covariate is post. That was by accident, was it? Must be the other way round like in the frequentist way I guess?
Best
Robin
Hi Robin,
Yes you are right, my apologies!
Cheers
Johnny
Thank you very much! Then I feel quite comfortable adding some bayesian analysis to my report =)