Asymptotic normality of the LkL_k-error of the Grenander estimator

preprint2006arXivOpen access

Abstract

We investigate the limit behavior of the LkL_k-distance between a decreasing density ff and its nonparametric maximum likelihood estimator f^n\hat{f}_n for k≥1k\geq1. Due to the inconsistency of f^n\hat{f}_n at zero, the case k=2.5k=2.5 turns out to be a kind of transition point. We extend asymptotic normality of the L1L_1-distance to the LkL_k-distance for 1≤k<2.51\leq k<2.5, and obtain the analogous limiting result for a modification of the LkL_k-distance for k≥2.5k\geq2.5. Since the L1L_1-distance is the area between ff and f^n\hat{f}_n, which is also the area between the inverse gg of ff and the more tractable inverse UnU_n of f^n\hat{f}_n, the problem can be reduced immediately to deriving asymptotic normality of the L1L_1-distance between UnU_n and gg. Although we lose this easy correspondence for k>1k>1, we show that the LkL_k-distance between ff and f^n\hat{f}_n is asymptotically equivalent to the LkL_k-distance between UnU_n and gg.

Explore this paper’s authors, topics and related work.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Reviews 0

Write a reviewWrite

No reviews yet.

Discussion 0

Add a commentWrite

No comments yet.