Penalized Regression Computation and Simulations

Abstract

In the specific realm of high-dimensional data, conventional regression models either have no solutions or, at best, have solutions that are highly sensitive to even small changes in the data. Thus, "penalized" regression methods are used, which add a constraint on the values fit by the model. Using various penalized regression models, the purpose of this study was to develop a cross-comparison of the different methods on simulated data that will identify situations (for sparse-predictor data) in which one method is preferred over another.