Abstract

We propose a new method to select the tuning parameter in lasso regression. Unlike the previous proposals, the method is iterative and thus it is particularly efficient when multiple tuning parameters have to be selected. The method also applies to more general regression frameworks, such as generalized linear models with non-normal responses. Simulation studies show our proposal performs well, and most of times, better when compared with the traditional Bayesian Information Criterion and Cross validation.
Lingua originaleEnglish
Pagine133-136
Numero di pagine4
Stato di pubblicazionePublished - 2016

Fingerprint Entra nei temi di ricerca di 'Tuning parameter selection in LASSO regression'. Insieme formano una fingerprint unica.

Cita questo