I’m trying to visually explore the effects of an interaction term in a linear (Bernoulli distributed logit link) model. My idea was to was plot the posterior predictions for this model two times, 1) as normal and 2) with the interaction term set to zero (or something else), and then compare the plots.
Is there a way to manipulate the posterior draws for a model’s parameters after the model has been fitted? Does this make sense?
I’m working in R, using brms to interface with Stan.