I'd like to suggest the addition of a particular variable selection method for Regression called “NONNEGATIVE GARROTE” method (Breiman, 1995)[1]. It is similar to LASSO and RIDGE REGRESSION, but slightly different. NONNEGATIVE GARROTE method performs variable selection based on the contraints b_i > 0 for all i.
Please see the following reading material for NONNEGATIVE GARROTE:
https://www.jstor.org/stable/1269730?seq=1
https://link.springer.com/chapter/10.1007/978-3-642-25832-9_9
[1] Breiman, Leo. “Better Subset Regression Using the Nonnegative Garrote.” Technometrics, vol. 37, no. 4, 1995, pp. 373–384. JSTOR, www.jstor.org/stable/1269730. Accessed 30 Nov. 2020.
It would be invaluable to have this method as an available variable selection option in PROC LOGISTIC, PROC REG, or PROC DMREG.
