Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Best way to bring R character string into Stata as global

    Hi, I'm hoping to run R's lasso (glmnet package) on a set of variables from Stata, then bring the *names* of the subset variables into Stata as a global. I'm hoping for advice on strategy.

    I've often before run .R code from Stata using vanilla --- I would export a dataset from Stata, then run the .R code on it using vanilla, then produce a .dta file at the end of the .R code, and then use that .dta file in the latter part of the Stata .do file.

    However, this time I want to import a set of character strings (4 of them, 25-75 characters long) produced by R into Stata, not as data but as macros. I'm aware that I could export the character string vector as a 1-variable dataset into Stata, and then probably loop through the variable values to create a macro of the names... but that seems quite round-about.

    Using rcall seems likely to be the best option? I assume that I could load glmnet and run lasso using rcall, after creating R matrices out of Stata matrices... but I'm having trouble doing that first step. I get the error "too many numeric literals" when try to save a Stata matrix as an R matrix, as I note at the end of this (error-relevant) thread here: https://www.statalist.org/forums/for...ls-r-130/page2

    Thanks in advance for any tips on best use of R-related code/commands!

  • #2
    Very very late to respond to this, but anyway...

    It's likely that the lasso routines in our Stata package LASSOPACK would mean you could do it all in Stata. We benchmarked/replicated vs glmnet. And selected variables are saved as macros.

    --Mark

    Comment

    Working...
    X