Feasible Generalized Least Squares Using Machine Learning

25 Pages Posted: 10 May 2017 Last revised: 6 Feb 2018

See all articles by Steve Miller

Steve Miller

University of Colorado at Boulder

Richard Startz


Date Written: February 6, 2018


In the presence of heteroskedastic errors, regression using Feasible Generalized Least Squares (FGLS) offers potential efficiency gains over Ordinary Least Squares (OLS). However, FGLS adoption remains limited, in part because the form of heteroskedasticity may be misspecified. We investigate machine learning methods to address this concern, focusing on Support Vector Regression. Monte Carlo results indicate the resulting estimator and an accompanying standard error correction offer substantially improved precision, nominal coverage rates, and shorter confidence intervals than OLS with heteroskedasticity-consistent (HC3) standard errors. Reductions in root mean squared error are over 90% of those achievable when the form of heteroskedasticity is known.

Keywords: Heteroskedasticity, Support Vector Regression, Weighted Regression

JEL Classification: C13, C21

Suggested Citation

Miller, Steve and Startz, Richard, Feasible Generalized Least Squares Using Machine Learning (February 6, 2018). Available at SSRN: https://ssrn.com/abstract=2966194 or http://dx.doi.org/10.2139/ssrn.2966194

Steve Miller (Contact Author)

University of Colorado at Boulder ( email )

1070 Edinboro Drive
Boulder, CO CO 80309
United States

Richard Startz

UCSB ( email )

Department of Economics
University of California
Santa Barbara, CA 93106-9210
United States
805-893-2895 (Phone)

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
PlumX Metrics