How to see which coefficients are not 0 in cv.glmnet

Question

I did a lasso regression by using the glm.net package. I chosed my lambda via cross validation and have now the information that the optimal model has only 6 coefficients that are not zero.

How can i see which coefficients that exactly are ?

Answer 1

Since you don't provide any sample data here is a minimal example:

1. Generate some sample data

   set.seed(2017);
   x1 <- seq(1:100);
   x2 <- 2 * x1;
   y <- 3 * x1 + 6 * x2 + rnorm(100);
   

2. Fit the model using CV

   fit <- cv.glmnet(cbind(x1, x2), y);
   

3. Then coef(fit) gives the parameter estimates for different lambda values. We can extract parameter estimates for the lambda value that results in the smallest CV error with

   coef(fit, s = "lambda.min")
   #3 x 1 sparse Matrix of class "dgCMatrix"
   #                       1
   #(Intercept) 2.439590e+01
   #x1          1.451704e+01
   #x2          5.723395e-16