OHDSI Home | Forums | Wiki | Github

"No regularization will be applied to the treatment variable". How to exclude that variable from being regularized?

Hello.

In the Book of OHDSI, in the section about Population-Level Estimation we can read:

“If we choose to include all covariates in the outcome model, it may make sense to use regularization when fitting the model if there are many covariates.
Note that no regularization will be applied to the treatment variable to allow for unbiased estimation”

But I have never seen this option in the common R packages.
What R package allows us to regularize the covariates while excluding one of them? How?
For example, if we want to perform a survival regression or a regression with random effects, with lasso regularization.

This done through the Cyclops package. See documentation of the createPrior() function.

t