I’d like to transform my response variable when using pp_check.brmsfit(). According to the documentation of pp_check.brmsfit(), arguments may be passed to predict.brmsfit() via “…” (ellipsis). Since predict.brmsfit() has the argument “transform”, I thought I could use the ellipsis to transform my response for pp_check.brmsfit(). However, this gives me the warning message “The following arguments were unrecognized and ignored: transform”. So my question is: What is the right way to transform the response for pp_check.brmsfit()?
This should actually still be working I guess. The warning comes from bayesplot::pp_check to which the argument is also passed and can savely be ignored.
Alright, thanks. However, when inspecting the resulting plot, the original data (thick black line) is not transformed while the predicted data (thin gray lines) is indeed transformed (which I guess makes sense considering that predict.brmsfit() is only responsible for generating predicted data). Is there a way to transform the original data as well?
I would just manually create yrep via posterior_predict, transform both y and yrep and pass it to the ppc functions of the bayesplot package directly.
Perfect, thanks.