Select a VCMoE bandwidth by K-fold cross-validation

Select a VCMoE bandwidth by K-fold cross-validation

Usage

vcmoe_select_bandwidth(
  formula,
  data,
  u,
  k = 2L,
  family = "gaussian",
  bandwidth_grid = NULL,
  folds = 5L,
  u_grid = NULL,
  control = list(),
  label = "align",
  parameterization = "a1_epanechnikov_scaled",
  u_scale = c("unit", "none"),
  seed = NULL,
  refit = TRUE,
  engine = c("local_grid_em", "joint_path_em")
)

Arguments

formula

A formula of the form y ~ expert_terms | gating_terms.

data

A data frame.

u

Continuous index column name or numeric vector.

k

Number of mixture components. Values from 2 through 10 are supported.

family

Model family. "gaussian", "binomial", and "negative-binomial" are supported.

bandwidth_grid

Candidate bandwidth values. If NULL, uses multiples of the default bandwidth.

folds

Number of random cross-validation folds.

u_grid

Grid where coefficient functions are estimated.

control

Named list passed to the selected fitting engine.

label

Label strategy passed to the selected fitting engine.

parameterization

Estimator convention passed to the selected fitting engine.

u_scale

u scaling strategy passed to the selected fitting engine.

seed

Optional random seed for fold assignment and, when control$seed is absent, deterministic CV refits.

refit

Whether to refit the final model on all data using the selected bandwidth.

engine

Fitting engine used for every cross-validation fold and the optional final refit. The default "local_grid_em" preserves the 0.1.0 behavior; "joint_path_em" uses the joint-path engine of vcmoe_fit().

Value

An object of class vcmoe_bandwidth_selection.