Bootstrap Bias Adjusted Estimators of Beta Distribution Parameters

Abstract

The standard beta distribution is one of the few well-studied distributions with continuous support on (0, 1] . In this paper, we propose a bootstrap bias corrected estimator of the standard beta shape parameters to improve on the traditional Method of Moments Estimators (MME) and Maximum Likelihood Estimators (MLE). Through empirical studies we show that the MME's have smaller bias but very large Mean Squared Error (MSE) as compared to the MLE's, while the MLE's have larger bias but substantially lower MSE. The bootstrap bias corrected estimators give substantial improvement in bias and MSE over both the MME and MLE.