# using caret and glmnet for variable selection

Im using the caret and glmnet package for variable selection. I only want to find the best model and the coefficients and use them for a different model. Please help me understand the differences between caret and glmnet. Here's an example:

using state data:

data(state)
statedata <- data.frame(state.x77,row.names=state.abb,check.names=T)
statedata <- scale(statedata)


in caret:

library(caret)
fitControl <- trainControl(## 10-fold CV
method = "repeatedcv",
number = 5,
repeats = 5)

lassoFit1 <- train(Life.Exp ~ . , data = statedata,
method = "glmnet",
trControl = fitControl)


I get alpha=1 and lambda=0.1 as best values. But how do I get the final best model? I tried

coef(lassoFit1$finalModel)  but I get a whole list of models. coef(lassoFit1$bestTune)
coef(lassoFit1$bestTune$.lambda)


both return NULL.

Here's how I do it in glmnet:

library(glmnet)
cvfit <- cv.glmnet(x=statedata[,c(1:3,5,6,7,8)],y=statedata[,4])


When I type

coef(cvfit, s="lambda.min")


I get the coefficients. Is this the best model? It looks like glmnet does test more lambdas in the cross validation, is that true? Does caret or glmnet lead to a better model?

How do I manage to extrage the best final model from caret and glmnet and plug it in a cox hazard model for example?

thanks

• you may want to try predictors(lassoFit1), which lists predictors used in the final model. – David Ren Feb 12 '16 at 0:32