Current browse context:
stat
Change to browse by:
References & Citations
Statistics > Methodology
Title: Estimating the error variance in a high-dimensional linear model
(Submitted on 6 Dec 2017 (v1), last revised 19 Jul 2019 (this version, v3))
Abstract: The lasso has been studied extensively as a tool for estimating the coefficient vector in the high-dimensional linear model; however, considerably less is known about estimating the error variance in this context. In this paper, we propose the natural lasso estimator for the error variance, which maximizes a penalized likelihood objective. A key aspect of the natural lasso is that the likelihood is expressed in terms of the natural parameterization of the multiparameter exponential family of a Gaussian with unknown mean and variance. The result is a remarkably simple estimator of the error variance with provably good performance in terms of mean squared error. These theoretical results do not require placing any assumptions on the design matrix or the true regression coefficients. We also propose a companion estimator, called the organic lasso, which theoretically does not require tuning of the regularization parameter. Both estimators do well empirically compared to preexisting methods, especially in settings where successful recovery of the true support of the coefficient vector is hard. Finally, we show that existing methods can do well under fewer assumptions than previously known, thus providing a fuller story about the problem of estimating the error variance in high-dimensional linear models.
Submission history
From: Guo Yu [view email][v1] Wed, 6 Dec 2017 21:16:35 GMT (52kb,D)
[v2] Wed, 17 Jan 2018 18:06:46 GMT (52kb,D)
[v3] Fri, 19 Jul 2019 09:15:05 GMT (94kb,D)
Link back to: arXiv, form interface, contact.