StepPenal {stepPenal}

Stepwise forward variable selection using penalized regression.

Description

Stepwise forward variable selection based on the combination of L1 and L0 penalties. The optimization is done using the "BFGS" method in stats::optim

Usage

StepPenal(Data, lamda, w, standardize = TRUE)

Arguments

Data

should have the following structure: the first column must be the binary response variable y.

lamda

the tuning penalty parameter

w

the weight parameter for the sum (1-w)L0+ wL1

standardize

Logical flag for the predictors' standardization, prior to fitting the model. Default is standardize=TRUE

Details

lamda and w parameters need to be tuned by cross-Validation using stepPenal::tuneParam

Value

a list with the shrinked coefficients and the names of the selected variables, i.e those variables with an estimated coefficient different from zero. It also returns the value of the objective function, evaluated for the values of the coefficients.

References

Vradi E, Brannath W, Jaki T, Vonk R. Model selection based on combined penalties for biomarker identification. Journal of biopharmaceutical statistics. 2018 Jul 4;28(4):735-49.

See Also

optim

Examples

# use the StepPenal function on a simulated dataset, with given lamda and w.

set.seed(14)
beta    <- c(3, 2, -1.6, -1)
noise   <- 5
simData <- SimData(N=100, beta=beta, noise=noise, corr=FALSE)
## Not run: 
before <- Sys.time()
stepPenal<- StepPenal(Data=simData, lamda=1.5, w=0.3)
after <- Sys.time()
after-before

(varstepPenal<- stepPenal$coeffP)

## End(Not run)