How do HCCMEs perform in small samples?

Authors

MEHMET ORHAN
ESRA ŞİMŞEK

DOI

10.3906/elk-1207-23

Abstract

The purpose of this paper is to explore the performances of prominent and popular heteroscedasticity-consistent covariance matrix estimators (HCCMEs) in small samples. The HCCMEs are selected from the literature and their ability to estimate the true covariance matrix of the coefficient's vector is evaluated through simulation runs. We calculate the percentage difference between the expected value of the HCCME and the true covariance matrix to set a convenient stage to make the comparisons under several different regression settings. The main contribution of the paper is the inclusion of the HCCMEs that have been introduced into the literature recently. We report the performances of the HCCMEs under different settings of the covariates and error term variances. We let the covariates follow uniform, normal, Student's t, and Cauchy distributions and tailor the error term variances to increase gradually.

Keywords

Heteroscedasticity-consistent covariance matrix estimator, minimum covariance determinant, Monte Carlo simulation

First Page

412

Last Page

424

Recommended Citation

ORHAN, MEHMET and ŞİMŞEK, ESRA (2014) "How do HCCMEs perform in small samples?," Turkish Journal of Electrical Engineering and Computer Sciences: Vol. 22: No. 2, Article 14. https://doi.org/10.3906/elk-1207-23
Available at: https://journals.tubitak.gov.tr/elektrik/vol22/iss2/14