Function for the total number of parameters

[avehtari]

loo package glossary and CV-FAQ mention that the effective number of parameters p_loo can be compared to the total number of parameters p. I will clarify in these, that the total number of parameters refer to data model parameters and not prior parameters.

It would be useful to have a function in brms that would provide the total number of parameters. It is possible to get a list of parameters and their dimensionality with posterior_summary(fit). The variable names in this output have certain structure. If I have understood correctly, at least the variables starting as following are data model parameters

b_
bs_
sigma
Intercept
r_
s_

the variables starting as following are hyperparameters

sd_
cor_
sds_

and the variables starting as following are due to non-centered parameterization

z_

Thus, it seems it might be possible to get the total number of parameters (excluding hyperparameters) from the currently available information.

EDIT: added z_