Forthcoming, Pacific Economic Review
20 Pages Posted: 11 Aug 2013 Last revised: 24 Aug 2013
Date Written: August 23, 2013
This paper compares a nonparametric generalized least squares (NPGLS) estimator to parametric feasible GLS (FGLS) and variants of heteroscedasticity robust standard error estimators (HRSEs) in an applied setting. Given myriad alternative HRSEs, a clear consensus on which version to use does not exist, and using alternative HRSEs can result in differing statistical conclusions. Further, FGLS is prone to model misspecification of the the scedastic function. Here we apply these various approaches to handling heteroscedasticity to data on professor rankings obtained from ratemyprofessor.com and find not only heteroscedasticity to be prevalent but NPGLS and FGLS provide different insights and the statistical significance of key variables varies across the seven versions of HRSEs.
Keywords: Heteroscedasticity, nonparametric, generalized least squares
JEL Classification: C13, C14
Suggested Citation: Suggested Citation
O'Hara, Michael and Parmeter, Christopher, Nonparametric Generalized Least Squares in Applied Regression Analysis (August 23, 2013). Forthcoming, Pacific Economic Review. Available at SSRN: https://ssrn.com/abstract=2308126 or http://dx.doi.org/10.2139/ssrn.2308126