cancelPenalty: Remove penalty from a penalized latent variable model

In bozenne/lava.penalty: Penalized Latent Variable Models

Description

Remove one or several penalties from a penalized latent variable model

Usage

cancelPenalty(x, ...)

cancelPenalty(x, ...) <- value

## S3 method for class 'plvm'
cancelPenalty(x, ..., value)

## S3 replacement method for class 'plvm'
cancelPenalty(x, ...) <- value

## S3 replacement method for class 'penaltyL12'
cancelPenalty(x, extraParameter, rm.lasso = TRUE,
  rm.ridge = TRUE, rm.groupLasso = TRUE) <- value

Arguments

x	plvm-object
...	additional arguments to be passed to lower level functions
value	the penalty that should be removed
extraParameter	external parameters
rm.lasso	should lasso penalties be removed?
rm.ridge	should ridge penalties be removed?
rm.groupLasso	should group lasso penalties be removed?

Details

Since lavaReduce::initVar_links does not work for the external parameters, they have to be handled separately.

Examples

## lasso
m <- lvm()
m <- regression(m, x=paste0("x",1:10),y="y")
pm <- penalize(m)

cancelPenalty(pm, value = "y~x5")
cancelPenalty(pm) <- y~x1
cancelPenalty(pm) <- c("y~x2","y~x3")
cancelPenalty(pm) <- penalty(pm, no.ridge = TRUE)$link
pm

## group lasso
m <- regression(m, x=paste0("x",1:10),y="y")
categorical(m, K = 3, labels = 1:3) <- ~x1
pm <- penalize(m)
pm
cancelPenalty(pm) <- "y~x2"
pm
cancelPenalty(pm) <- "y~x12"
pm
cancelPenalty(pm) <- penalty(pm, no.lasso = TRUE, no.ridge = TRUE)$link
pm

bozenne/lava.penalty documentation built on May 13, 2019, 1:41 a.m.