step.lmr: Stepwise model selection for a robust linear model

In harrysouthworth/margarita: Support functions for texmex

Description

Stepwise model selection for a robust linear model

Usage

step.lmr(
  object,
  scope,
  target = "RFPE",
  direction = "backward",
  trace = TRUE,
  steps = 1000,
  k = 2,
  cores = NULL,
  ...
)

Arguments

object	An object fit by lmr.
scope	The lowest model to consider.
target	Character string giving the method for selecting models. Defaults to "RFPE", the only current alternative being "AIC".
direction	What direction to take steps in. Only "backwards" is implemented.
trace	Whether or not to report progress. Defaults to trace=TRUE.
steps	The maximum number of steps to take.
k	The penalty factor to be used if AIC is being used as the target. Defaults to k=2.
cores	The number of cores to use when running drop1.lmr in parallel. Defaults to cores=NULL and the function will try to guess how many cores to use.
...	Not used (for compatibility with step).

Details

The function uses robust finite prediction error to decide when to stop the selection process. In principle, other approaches could be implemented.

harrysouthworth/margarita documentation built on Aug. 19, 2021, 5 a.m.