Panel Data模型设定的新思路①
———固定效应与随机效应的统一
张红星1、2　贾彦东2
(1·云南财贸学院; 2·西南财经大学统计学院
【摘要】经典Panel Data模型研究中一直存在着固定效应与随机效应的判断与
争论问题,这种模型设定形式的不准确常常导致模型参数估计的无效性以及一维
(one-way)与二维(two-way)误差成分模型的混淆。在此,本文提出建立一种新
的同时囊括随机与固定两种效应的误差成分一般模型,其中一维(one-way)情况
的模型仅为二维(two-way)模型的特例模型,单一的随机效应或固定效应模型亦
为其特殊情况的一种。在这种一般误差成分模型的基础上,我们力图将有关Panel
Data模型的讨论纳入到一个更加一般和统一的分析构架中予以研究。
关键词　Panel Data模型　方差成分一般模型　随机效应　固定效应
中图分类号　F224·0　　文献标识码　A
The New Idea of Panel Data Model
　　Arstract:The debate between fixed effect and random effect has been haunting
in Panel Data analysis since its origin·Discusses afterwards which try to make the
difference are clear on behalf of the involved applied researchers, but the distinc-
tion still baffles the followers·In this paper we propose a new approach to specify
the panel data model which makes both fixed effect and random effect as its special
cases, and also, the hierarchy among one-way and two-way is revealed clearly un-
der our approach·The most important implication is that it is possible to proceed
the analysis of panel data model in a unified framework·
Key words:Panel Data model, Fixed effect; Random Effect; Double Effect;
One-way Model, Two-way Model, General Model