Undo standardization after fit with brms

Hello Stan forum,

a small question, that many of you most likely have come across already.

After standardization of data and model fit I would like to rescale the posterior back to the original non-standardized unit for better interpretation / presentation.

What is your best practice to perform it on a brms fit object?

I understand the considerations behind the design decisions not to include this function into brms: https://github.com/paul-buerkner/brms/issues/440

I usually use „standardize“ from the effectsize package (https://easystats.github.io/effectsize). Which saves the used standardization values (center, scale) per parameter in an attribute with the data which is also in the later brms fit object. So maybe there is an elegant way to access this afterwards and still use the conditional_effects plots from brms?

Is there a tidybayse / tidyverse approach you know of? I could not find anything in that direction so far.

Thanks for your ideas.

Kruschke covered this topic for several single-level models in his text (http://doingbayesiandataanalysis.blogspot.com/), chapters 16 and 17, I believe.

Thanks for the hint to this more workflow related question. The book is orderd at the library and I have not checked your brms version of it yet, but I will do now.

The center and scale value can be retrived via the code below after standardization with the effectsize package. This should also work with the brms object later on:

attr(data_stan$parameter_name, "center")
attr(data_stan$parameter_name, "scale")
1 Like