Bug in brms bayes_factor for model with me() term?

Ad formatting: you can use triple backticks (```) to mark code/output blocks, I edited the post to show that.

I have little direct experience with Bayes factors, but many people I trust consider them fragile. You might as well be encountering one of the fragile cases. Measurement error models can generally be quite tricky in and of themselves and I am not sure Bayes factor comparison can even be made meaningful between a model with an without a measurement error term - your model gained many dimensions which you need to integrate over.

Checking the stability of the estimate by fitting the model multiple times and computing multiple bayes factors as mentioned at Bayes factor using brms might be worthwhile.

Tagging @Henrik_Singmann for second opinion (if he’s got time to spare).