Sorry that I confused with “stacking” and “pseudo BMA”.
I see, it is about combining predictive distributions.
That means, I need to have “y_rep” in generated quantities of my Stan model, and I need to fit the models again, but this time using “weighted” number of iterations, and then combine all “y_rep” at the end. And this is my combined posterior distribution for “y_rep”, is that right? Is this what you explained here (Implementing Model Averaging), (but for rstanarm mdoels) right?
Thank you again.