I have a mixed-effects regression model with a binary response variable Yes-No. The predictors are all binary as well. It turns out that when I calculate the Bayes factors for each parameter, the results change based on the reference level of the response variable and the predictors. When the reference level is Yes, one of the predictors gets a Bayes Factor of 0.14, when I change it to No the Bayes factor of the same predictor is < 1000.

How is this possible? Is this the way it is supposed to be?