dlsr-package: dlsr: Double Lasso Selection

Description Details Author(s) References


The doubleLassoSelect function is an alternative implementation of Double Lasso Selection on OLS. This implementation is based primarily on package glmet and a paper of O. Urminsky, C. Hansen, and V. Chernozhukov (working draft), as listed in the reference.

For original implementation, please see hdm package by Chernozhukov, Hansen, and Spindler. About the mathematical details of the method, please refer to the original papers listed in the reference section.


Compared to rlassoEffects function in the original hdm, doubleLassoSelect function in dlsr provides an alternative implementation of this specific method with the following benefits:

  1. The doubleLassoSelect function accepts character vectors as variable input, instead of matrix indices or logical vectors. This improves code readability and facilitates batch implementation with external data source such as a csv.

  2. It Supports interaction terms as variable input. The doubleLassoSelect handles the matrix expansion for you.

  3. Instead of the result of a linear model, the doubleLassoSelect outputs a data frame (data.table) with the selected variables. This provides users more flexibility in subsequent operations, for example, applying the selected result further in a latent class model.


Maintainer: Chih-Yu Chiang chihyuchiang@uchicago.edu


ChihYuChiang/dlsr documentation built on May 30, 2019, 10:45 p.m.