In a multivariate linear regression model with $p>1$ covariates, implementation of penalization techniques often implies a preliminary univariate standardization step. Although this prevents scale effects on the covariates selection procedure, possible dependence structures can be disrupted, leading to wrong results. This is particularly challenging in high-dimensional settings where $p \geq n$. In this paper, we analyze the standardization effect on the LASSO for different dependence-scales contexts by means of an extensive simulation study. Two distinct objectives are pursued: adequate covariate selection and proper predictive capability. Additionally, its behavior is compared with the one of some well-known or innovative competitors. This comparison is also extended to three real datasets facing different dependence-scales patterns. Eventually, we conclude with discussion and guidelines on the most suitable methodology for each case in terms of covariates selection or prediction.
翻译:暂无翻译