5
$\begingroup$

The Belloni et al. 2014 paper develops the theory of Lasso in post-double selection. There are two commands that appear to me to perform this function: $dsregress$ and $pdslasso$.

The $dsregress$ command is just "double-selection" while $pdslasso$ is "post-double-selection". Are these commands methodologically equivalent?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

Based on what is written in the Stata 17 Manual pp 60-66 and documentation for pdslasso, they are exactly the same (although dsregress has richer set of options).

  1. Both documentations reference Belloni et al. 2014 paper
  2. Both documentations show basically the same model (Stata manual transposes beta vector instead of x matrix but once you go through the math the expressions are equivalent as far as I can see).

Next, this would not be the first time when there already was user written package for Stata, and later when new version of Stata is released developer just add renamed version of that package. Stata is developed for profit, so I guess it makes sense to do things twice so they can sell users the "new Stata x.x" with "new" features like lasso.

Stata added lasso models, including dsregress close to the end of 2019 in Stata 16, the pdlasso package was out in late 2018/early 2019.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.