第 29 卷 第 1 期
2 0 1 2 年 3 月
经 济 数 学
JOURNAL OF QUANTITATIVE ECONOMICS
Vol .29 ,No .1
Mar .2 0 1 2
有限总体中基于广义 Liu 估计的预测及其小样本性质
倡
黄介武
1 ,2
(1 .贵州民族学院 理学院 ,贵州 贵阳 550025 ; 2 .重庆大学 数学与统计学院 ,重庆 401331)
摘 要 在有限总体中提出了一类基于广义 Liu 估计的新的预测 ,得到了基于广义 Liu 估计的预测在
预测均方误差意义下优于最优线性无偏预测的充要条件 ,并通过实例对理论成果进行了进一步的说明 .
关键词 有限总体 ;预测均方误差 ;最优线性无偏预测 ;广义 Liu 估计
中图分类号 O 212 .4 文献标识码 A
A Predictor Based on the Generalized Liu Estimator
in Finite Populations and Its Small‐sample Properties
HUAMG Jie‐wu
1 ,2
(1 .Colle
g
e o
f
Science ,Guiz hou Universit
y f
or N ationalities ,Guiz hou ,Gui
y
ang 550025 ;
2 .Colle
g
e o
f
M athematics and S tatistics ,Chon
gq
ing Universit
y
,Chon
gq
ing 401331)
Abstract
This paper proposed a new predictor based on the generalized Liu estimator in finite populations .The necessa‐
ry and sufficient conditions for the superiority of the predictor based on the generalized Liu estimator over the best linear unbi‐
ased predictor in the prediction mean squared error sense were derived .Furthermore ,a numerical example was given to illus‐
trate some of the theoretical results .
Key words
finite populations ;
p
rediction mean squared error ;best linear unbiased predictor ,
g
eneralized Liu estimator
1 Introduction
The problem of prediction in finite populations
received considerable attention in the past several
decades .As we can see ,most of works in the exist‐
ing literatures are in the context of linear unbiased
p
rediction , such as Hendserson
[1]
, Perieira and
Kodrigues
[2]
,Bolfarine et al .
[3]
,Yu and He
[4]
and
Rueda et al .
[5]
. However ,the prediction mean
squared error of the best linear unbiased predictor
(BLUP ) will become inflated when there exists
multicollinearity among the explanatory variables
and the results of regression will be often unac‐
ceptable .In this case , some prediction methods
have been developed to improve the best linear un‐
biased predictor ,one of these methods is the bi‐
ased prediction method .Hoerl and Kennard
[6]
p
ro‐
p
osed the so‐called ridge regression method which
isunaffected by multicollinearity among the many
independentvariables . Wang
[7]
p
roposed adaptive
ridge‐type predictors in finite population .
Our primary aim in this paper is to introduce a
new predictor based on the generalized Liu estima‐
tor
[8]
to combat the multicollinearity . The pro‐
p
osed predictor is biased ,and we discuss its supe‐
倡
收稿日期 :2011‐12‐28
基金项目 :贵州省科学技术厅 、贵州民族学院科学技术联合基金(黔科合 J 字 LKM [2011]09 号 )
作者简介 :黄介武 (1977 — ) ,男 ,湖南益阳人 ,副教授 ,博士研究生
E‐mail :huangjiewu@ sohu .com