Nonparametric regression estimators for length biased data
Nonparametric regression estimators for length biased data
复制标题
DOI：
10.1016/s0378-3758(00)00092-6
复制
发表时间：
2000-08-15
影响因子：
0.9
通讯作者：
Alcalá, JT
中科院分区：
文献类型：
--
作者：
Cristóbal, JA;Alcalá, JT
普通的核回归在应用于长度有偏数据抽样时并不令人满意。本文利用修正的局部多项式提出了回归函数的几种估计，并研究了它们的渐近最优带宽和渐近均方误差。我们还分析的情况下，我们有两种类型的样本，第一个长度有偏的数据和第二个直接从人口：在这种情况下，我们计算的渐近偏差和方差的估计下三种替代方法。最后，进行了模拟，以比较这些估计与有限样本的行为。(C)2000 Elsevier Science B.V.保留所有权利。
Ordinary kernel regression is not satisfactory when applied to length biased data sampling. In this paper, we propose several estimators of the regression function by way of modified local polynomials, and study their asymptotic optimal bandwidth and asymptotic mean squared error. We also analyze the situation where we have two types of samples, the first with length biased data and the second obtained directly from the population: in such cases, we calculate the asymptotic bias and variance of the resulting estimators under three alternative approaches. Finally, a simulation is carried out to compare the behavior of these estimators with finite samples. (C) 2000 Elsevier Science B.V. All rights reserved.