Interpreting Bayesian ANCOVA results
I’m looking for help interpreting evidence for a one-way Bayesian ANCOVA looking at heart rate response to emotional stimuli, screenshots attached. Between-subjects factor is Group (clinical/control), and HR response (incongruentHRaverage) is the dependent variable (DV). Baseline HR is not independent from Group, and is highly correlated with the DV, thus is kept in as a covariate recognising that it is part of the total effect. We didn’t use change scores because there is evidence that it is problematic (https://pubmed.ncbi.nlm.nih.gov/34100077/). Age and sex are covariates of no interest (ie, included in the null model). The interaction between group and baseline is also included.
Just wondering - baseline is the winning model, but how would you interpret the effect of group - comparatively tiny, but still a Bayes factor of 89, but also with the complication of an interaction between group and baseline measure…
Any thoughts/advice appreciated, thank you!
Comments
The results will make more sense if you compare not to the null model, but to the best model. It looks like the "group" model and the null model both perform very poorly. The other three models are fairly close...