brm

Model fitting wrappers.

This module contains the brms::brm() wrapper used by brmspy.brms.fit() / brmspy.brms.brm().

Notes

This code executes inside the worker process (the process that hosts the embedded R session).

Classes

Functions

brm(formula, data, priors=None, family='gaussian', sample_prior='no', sample=True, backend='cmdstanr', formula_args=None, cores=2, *, return_idata=True, **brm_args)

brm(formula: FormulaConstruct | ProxyListSexpVector | str, data: dict | pd.DataFrame, priors: Sequence[PriorSpec] | None = ..., family: str | ListSexpVector | None = ..., sample_prior: str = ..., sample: bool = ..., backend: str = ..., formula_args: dict | None = ..., cores: int | None = ..., *, return_idata: Literal[True] = True, **brm_args: Any) -> FitResult

brm(formula: FormulaConstruct | ProxyListSexpVector | str, data: dict | pd.DataFrame, priors: Sequence[PriorSpec] | None = ..., family: str | ListSexpVector | None = ..., sample_prior: str = ..., sample: bool = ..., backend: str = ..., formula_args: dict | None = ..., cores: int | None = ..., *, return_idata: Literal[False], **brm_args: Any) -> ProxyListSexpVector

Fit a Bayesian regression model with brms.

This is a thin wrapper around R brms::brm() that returns a structured FitResult (including an ArviZ InferenceData).

Parameters:

Name	Type	Description	Default
formula	str or FormulaConstruct	Model formula. Accepts a plain brms formula string (e.g. "y ~ x + (1|g)") or a composed formula created via brmspy.brms.bf() / brmspy.brms.lf() (typically imported as from brmspy.brms import bf, lf).	required
data	dict or DataFrame	Model data.	required
priors	Sequence[PriorSpec] or None	Optional prior specifications created via brmspy.brms.prior().	None
family	str or ListSexpVector or None	brms family specification (e.g. "gaussian", "poisson").	"gaussian"
sample_prior	str	Passed to brms. Common values: "no", "yes", "only".	"no"
sample	bool	If False, compile the model without sampling (brms empty=TRUE).	True
backend	str	Stan backend. Common values: "cmdstanr" or "rstan".	"cmdstanr"
formula_args	dict or None	Reserved for future use. Currently ignored.	None
cores	int or None	Number of cores for brms/cmdstanr.	2
return_idata	bool	When working with large datasets, you might not want the full idata. when False, you get the R object proxy which can be forwarded to posterior_epred or other functions	True
**brm_args		Additional keyword arguments passed to R brms::brm() (e.g. chains, iter, warmup, seed).	{}

Returns:

Type	Description
FitResult	Result object with idata (ArviZ InferenceData) and an underlying R handle.

See Also

brms::brm : R documentation

Warnings

Using cores <= 1 can be unstable in embedded R sessions and may crash the worker process. Prefer cores >= 2.

Examples:

from brmspy import brms

fit = brms.brm("y ~ x + (1|g)", data=df, family="gaussian", chains=4, cores=4)

fit.idata.posterior