Unified reciprocal LASSO estimation via least squares approximation

The primary goal of this article is to extend the reciprocal LASSO for applications binary and survival outcomes. We consider least squares approximation (LSA) as a solver problem. LSA general theoretical framework that includes generalized linear models, Cox regression, many others special cases. By applying regularization, we transfer original problem into an asymptotically equivalent While existing literature on has mostly focused our algorithm can be easily implemented likelihoods, providing flexible variable selection using penalties. To handle computational burden implementing resulting procedure, employ scalable stochastic search method called Simplified Shotgun Stochastic Search with Screening (S5), which easy implement, without requiring any sophisticated optimization package other than equation solver. examine effectiveness procedure through Monte Carlo simulations real data analyses.