Comparison of improved regression estimators with and without moments

Abstract

For estimating the coefficients in a linear regression model, the double k–class estimators are considered and the small disturbance asymptotic approximation for their density function is obtained. Then employing the criterion of concentration probability around the true parameter values, a comparison is made between the estimators possessing finite moments and the estimators having no finite moments.

Keywords:

• Concentration probability
• double k-class estimators
• small disturbance asymptotic distribution