Dummy / Reference variable in LASSO (group lasso)

Question

I am performing group lasso and need to double check if I include a dummy variable for the reference answer or not. For example:

original question : no (0), Yes (1), Unknown (9).

If I create 3 dummy variables, no (reference) (0/1), yes (0/1), and unknown (0/1), would I include all three in the group lasso or just the 2 (yes and unknown).

score 1 · Answer 1 · answered Mar 21 '23 at 22:45

1

It doesn't really matter because you are using regularization. Without regularization, you would have too many parameters, but it's not a problem for a regularized model. So you can use either. Including all the categories is a popular choice for regularized models though, as you don't need to decide which category to drop.

answered Mar 21 '23 at 22:45

Tim

138,066

For my logistic regression model, after regularization. Would I include all three parameters (no/yes/unk) or only the two yes/unk and leave the reference out? – Levi M Mar 22 '23 at 15:11
@LeviM as said above, it's up to you, both are ok. – Tim Mar 22 '23 at 15:13
Ok, thank you for confirming so quickly! – Levi M Mar 22 '23 at 15:14

Dummy / Reference variable in LASSO (group lasso)

1 Answers1