calcLambda: Estimate the regularization parameter

In bozenne/lava.penalty: Penalized Latent Variable Models

Description

Find the optimal regularization parameter according to a fit criterion

Usage

calcLambda(x, ...)

## S3 method for class 'plvmfit'
calcLambda(x, seq_lambda1, data.fit = x$data$model.frame,
  data.test = x$data$model.frame, warmUp = lava.options()$calcLambda$warmUp,
  fit = lava.options()$calcLambda$fit, trace = TRUE, ...)

Arguments

x	a plvmfit object
...	additional arguments - e.g. control argument for estimate.lvm
seq_lambda1	the sequence of penalisation paramaters to be used to define the sub model
data.fit	the data used to fit the model
data.test	the data used to test the model
warmUp	should the new model be initialized with the solution of the previous model (i.e. estimated for a different penalization parameter)
fit	criterion to decide of the optimal model to retain among the penalized models.
trace	shoud the execution of the function be traced

Examples

set.seed(10)
n <- 300
formula.lvm <- as.formula(paste0("Y~",paste(paste0("X",1:4), collapse = "+")))
mSim <- lvm(formula.lvm)
df.data <- sim(mSim,n)

lvm.model <- lvm(formula.lvm)
plvm.model <- penalize(lvm.model)

res <- estimate(plvm.model, data = df.data, increasing = FALSE, regularizationPath = TRUE)

perf <- calcLambda(res$regularizationPath, res$x, data.fit = df.data)
perf

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