lassomodel: Fits a lasso model and a lasso followed by a stepAIC...

In stepPenal: Stepwise Forward Variable Selection in Penalized Regression

Description

Fits a lasso model and a lasso followed by a stepAIC algorithm.

Usage

lassomodel(Data, standardize = TRUE, measure = c("deviance"),
  nfold = 5)

Arguments

Data	a data frame, as a first column should have the response variable y
standardize	Logical flag for variable standardization, prior to fitting the model. Default is standardize=TRUE. If variables are in the same units already, you might not wish to standardize.
measure	loss to use for cross-validation. measure="auc" is for two-class logistic regression only, and gives area under the ROC curve. measure="deviance", uses the deviance for logistic regression.
nfold	number of folds - default is 5. Although nfolds can be as large as the sample size (leave-one-out CV), it is not recommended for large datasets. Smallest value allowable is nfolds=3

Details

the function lassomodel is a wrapper function over the glmnet::glmnet. The parameter lambda is tuned by 10-fold cross-validation with the glmnet::cv.glmnet function. The selected lambda is the one that gives either the minimum deviance (measure="deviance") or the maximum auc (measure="auc") or minimum misclassification error (measure="class")

Value

a list with the coefficients in the final model for the lasso fit and also for the lasso followed by stepAIC.

See Also

glmnet

Examples

## Not run: 
set.seed(14)
beta    <- c(3, 2, -1.6, -4)
noise   <- 5
simData <- SimData(N=100, beta=beta, noise=noise, corr=FALSE)

lassofit <- lassomodel(Data=simData, measure="auc")
lassofit

lassofit2 <- lassomodel(Data=simData, measure="deviance")
lassofit2

## End(Not run)

stepPenal documentation built on May 1, 2019, 10:11 p.m.