How get coefs name from glmnet in R?

Question

I am doing lasso with glmnet in R. I have the lasso coefs out$beta

Input:

x1  0.5744207722
x2 -0.3575682570
x3 -0.0995794334

How can I extract the name of the variables?

Desired output: x1, x2,x3

What I tried

$
as.data.frame
as.matrix

but none of them convert the coefs into a n row 2 columns matrix. They remain as a vector, x1 0.5744207722.

I cannot do it manually. I have more than 1000 variables and more models to run. And I don't won't to write it on my disk, and read it back.

Did you look at the docs for `glmnet`? The very first example shows you how to pull out the coefficients for each variable using `coef()`. — Barker, Jan 25 '17 at 22:22
I want names `x1`, `x2` not coefs `0.5744207722`, `-0.3575682570` — Jill Clover, Jan 25 '17 at 22:23
Then `rownames(coef(fit))`. If that doesn't do what you want, give a reproducible example with inputs and desired outputs. — Barker, Jan 25 '17 at 22:25
If you try yourself, it would return `Error: $ operator is invalid for atomic vectors`. The question has been revised. — Jill Clover, Jan 25 '17 at 22:29
The reason is glmnet returns a vector rather than a matrix or data frame. — Jill Clover, Jan 25 '17 at 22:30
@alex23lemm Please move the comment to the answer section. Will accept it. Very useful. Very practical. — Jill Clover, Jan 25 '17 at 23:19

alex23lemm · Accepted Answer · 2017-01-26T16:21:21.230

Based on your input above it seems that you specified a glmnet call with a single lambda value.

In this particular case you can extract the names of the coefficients like this: names(out$beta[, 1][out$beta[, 1] != 0]). Please note that only the names of the non-zero betas are extracted, which makes sense when applying Lasso here since Lasso performs feature reduction.

A minimum reproducible example would be this: out <- glmnet(as.matrix(mtcars[-1]), mtcars[["mpg"]], lambda = 1)

How get coefs name from glmnet in R?

1 Answers1