人口研究 ›› 2011, Vol. 35 ›› Issue (6): 3-12.

• 论文 •    下一篇

对纵贯数据统计分析的认识

任强1谢宇2   

  1. 1 北京大学人口研究所,北京 100871; 2 密歇根大学社会学系,北京大学中国社会科学调查中心,北京 100871
  • 出版日期:2011-11-29 发布日期:2012-11-18
  • 通讯作者: renqiang@pku.edu.cn
  • 作者简介:1 北京大学人口研究所副教授;2 密歇根大学社会学系、北京大学中国社会科学调查中心教授

Statistical Analysis of Longitudinal Data

Ren Qiang1; Xie Yu2   

  1. 1 Institute of Population Research, Peking University, Beijing 100871; 2 Department of Sociology at the University of Michigan and the Institute of Social Science Survey at Peking University, Beijing 100871
  • Online:2011-11-29 Published:2012-11-18
  • About author:1 Associate Professor, Institute of Population Research, Peking University; 2 Professor, Department of Sociology at the University of Michigan and the Institute of Social Science Survey at Peking University

摘要: 在介绍了纵贯数据的设计思想及优缺点基础上,从统计方法的角度讨论了纵贯数据在社会科学中所能发挥的作用。纵贯数据的优点在于其可以帮助我们进行对总体异 质性的识别、对因果机制干预的研究、对因果效应的研究和对"状态"变换的研究。以一些基于纵贯数据的研究设计为实例,文章阐述了在研究中假设与数据紧密衔接的重要性,以及利用统计方法分析纵贯数据时需要考虑的要点。但由于存在着由人类和人类行为内在变异性导致的这一无法避免的根本性缺陷,纵贯数据并不能解决所有问题。因而在纵贯数据的辅助下,研究者需要对社会现象有更深入的理解,将其进行更合理的概念化,并加以更精准的数据分析。

关键词: 纵贯数据, 因果效应, 变异性, 异质性

Abstract: The paper introduces the basic ideas of design for longitudinal survey data and its advantages and shortcomings,and discusses the rationales for collecting longitudinal data from the statistical perspectives.Longitudinal data are informative because they enable identification of population heterogeneity,study of intervening causal mechanisms,study of causal effects,and study of state transitions.Special considerations in longitudinal settings are addressed,as well as the importance of hypotheses,illustrated with examples of study designs using longitudinal data.Longitudinal data are not perfect,because the most serious shortcomings come from the intrinsic variability of humans and human behaviors.Given such severe limitations,what researchers of social phenomena can do is to develop better understanding,better conceptualization,and better data analysis,aided by longitudinal data.

Keywords: