首页 | 本学科首页   官方微博 | 高级检索  
     检索      


On the sensitivity of pre-test estimators to covariance misspecification
Authors:Huaizhen Qin
Institution:a Department of Epidemiology and Biostatistics, Case Western Reserve University, School of Medicine, Cleveland, OH 44106, USA
b MADIS, Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100190, China
c Department of Management Sciences, City University of Hong Kong, Kowloon, Hong Kong
Abstract:In statistical and econometric practice it is not uncommon to find that regression parameter estimates obtained using estimated generalized least squares (EGLS) do not differ much from those obtained through ordinary least squares (OLS), even when the assumption of spherical errors is violated. To investigate if one could ignore non-spherical errors, and legitimately continue with OLS estimation under the non-spherical disturbance setting, Banerjee and Magnus (1999) developed statistics to measure the sensitivity of the OLS estimator to covariance misspecification. Wan et al. (2007) generalized this work by allowing for linear restrictions on the regression parameters. This paper extends the aforementioned studies by exploring the sensitivity of the equality restrictions pre-test estimator to covariance misspecification. We find that the pre-test estimators can be very sensitive to covariance misspecification, and the degree of sensitivity of the pre-test estimator often lies between that of its unrestricted and restricted components. In addition, robustness to non-normality is investigated. It is found that existing results remain valid if elliptically symmetric, instead of normal, errors are assumed.
Keywords:Linear restrictions  Non-spherical disturbances  Pre-test  Robustness
本文献已被 ScienceDirect 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号