Learn R Programming

brms (version 2.19.0)

get_refmodel.brmsfit: Projection Predictive Variable Selection: Get Reference Model

Description

The get_refmodel.brmsfit method can be used to create the reference model structure which is needed by the projpred package for performing a projection predictive variable selection. This method is called automatically when performing variable selection via varsel or cv_varsel, so you will rarely need to call it manually yourself.

Usage

get_refmodel.brmsfit(
  object,
  newdata = NULL,
  resp = NULL,
  cvfun = NULL,
  dis = NULL,
  latent = FALSE,
  brms_seed = NULL,
  ...
)

Value

A refmodel object to be used in conjunction with the

projpred package.

Arguments

object

An object of class brmsfit.

newdata

An optional data.frame for which to evaluate predictions. If NULL (default), the original data of the model is used. NA values within factors are interpreted as if all dummy variables of this factor are zero. This allows, for instance, to make predictions of the grand mean when using sum coding.

resp

Optional names of response variables. If specified, predictions are performed only for the specified response variables.

cvfun

Optional cross-validation function (see get_refmodel for details). If NULL (the default), cvfun is defined internally based on kfold.brmsfit.

dis

Passed to argument dis of init_refmodel, but leave this at NULL unless projpred complains about it.

latent

See argument latent of extend_family. Setting this to TRUE requires a projpred version >= 2.4.0.

brms_seed

A seed used to infer seeds for kfold.brmsfit and for sampling group-level effects for new levels (in multilevel models).

...

Further arguments passed to init_refmodel.

Details

Note that the extract_model_data function used internally by get_refmodel.brmsfit ignores arguments wrhs and orhs. This is relevant for predict.refmodel, for example.

Examples

Run this code
if (FALSE) {
# fit a simple model
fit <- brm(count ~ zAge + zBase * Trt,
           data = epilepsy, family = poisson())
summary(fit)

# The following code requires the 'projpred' package to be installed:
library(projpred)

# perform variable selection without cross-validation
vs <- varsel(fit)
summary(vs)
plot(vs)

# perform variable selection with cross-validation
cv_vs <- cv_varsel(fit)
summary(cv_vs)
plot(cv_vs)
}

Run the code above in your browser using DataLab