Again, this is not urgent at all, but rather a cosmetic feature request for something I encountered while working on shinybrms. And, similar to #1251, this might not be easy to resolve or at least it might not be worth the effort.
When using the cmdstanr backend, there are some redundant parameters which are "exported" by as.matrix(<brmsfit_object>) and launch_shinystan(<brmsfit_object>), for example. At least there is an inconsistency with the rstan backend.
Reprex:
data(bacteria, package = "MASS")
bfit_rstan <- brms::brm(
formula = y ~ 1 + week * trt + (1 | ID),
data = bacteria,
family = brms::bernoulli(),
chains = 2,
iter = 500,
seed = 1234
)
bfit_cmdstanr <- brms::brm(
formula = y ~ 1 + week * trt + (1 | ID),
data = bacteria,
family = brms::bernoulli(),
chains = 2,
iter = 500,
seed = 1234,
backend = "cmdstanr"
)
I agree, and already have a TODO in the code. I don't have time to work on this at the moment but would appreciate a PR if anyone has time to work on it.
Again, this is not urgent at all, but rather a cosmetic feature request for something I encountered while working on shinybrms. And, similar to #1251, this might not be easy to resolve or at least it might not be worth the effort.
When using the cmdstanr backend, there are some redundant parameters which are "exported" by
as.matrix(<brmsfit_object>)
andlaunch_shinystan(<brmsfit_object>)
, for example. At least there is an inconsistency with the rstan backend.Reprex:
Now
gives
and
gives
These parameters exported additionally when using the cmdstanr backend could be omitted, I think.
Session info:
CmdStan version: