A0483
Title: High-dimensional generalized penalized least squares
Authors: Ilias Chronopoulos - University of Essex (United Kingdom) [presenting]
Aikaterini Chrysikou - Kings College, University of London (United Kingdom)
George Kapetanios - Kings College London (United Kingdom)
Abstract: Inference is developed in high dimensional linear models with serially correlated errors. The Lasso estimator is examined under the assumption of a-mixing in the covariates and error processes. While the Lasso estimator performs poorly under such circumstances, it is estimated via GLS Lasso the parameters of interest and the asymptotic properties of the Lasso are extended under more general conditions. The theoretical results indicate that the non-asymptotic bounds for stationary dependent processes are sharper, while the rate of Lasso under general conditions appears slower as T,p. Further, debiasing methods are employed to perform inference uniformly on the parameters of interest. Monte Carlo results support the proposed estimator, as it has significant efficiency gains over traditional methods.
