glmnet.process.output: The glmnet model postprocessor

In cezden/lineaRutils: Utils for Linear Models

Description

The function is based on the results returned by glmnet_tidiers, with an optional application of results filtering using the structural.changes.col.matrix function.

Usage

glmnet.process.output(glmnet.res, nonzero.coeffs.only = FALSE,
  structural.filter.type = c("none", "initial", "final"), pseudoAIC = FALSE)

Arguments

glmnet.res	the result of the glmnet function
nonzero.coeffs.only	if TRUE only the non-zero model coefficients are returned in the coeffs field
structural.filter.type	the type of the steps filter, selecting the steps from the lambda path resulting in the change of the model structure: ("none"): all steps are returned (default) ("initial"): return initial step of each subpath consisting of the same variables ("final"): return final step of each subpath consisting of the same variables
pseudoAIC	(logical) should AIC be calculated for each model; Remark: the value being calculated is 2*df -2*loglike + 2*loglike(NULL), where df is an approximation of df returned by glmnet

Value

the list with following fields:

• steps: a data.frame with columns:

  • step

  • lambda

  • dev.ratio

  • df

  • dev

  • (optionally) pseudoAIC

• coeffs: a data.frame with columns:

  • step

  • term

  • estimate

cezden/lineaRutils documentation built on May 13, 2019, 3:07 p.m.