Maximum Spacing Estimates: A Generalization and Improvement on Maximum Likelihood Estimates I

摘要: Hoffmann-Jorgensen (1992) presents a fairly general framework for considering consistency and asymptotic normality of “M-estimates.” The most celebrated such estimate is the famous maximum likelihood (MLE), about which there has been considerable controversy due to fact that it may not even be consistent [see Le Cam (1990)]. Hoffmann-Jorgensen’s theory tries unify discussion inconsistency, but does extend domain MLE’s consistency. current paper was motivated by first trying understand why MLE inconsistent then alternatives might rectify inconsistency.