摘要
多层(嵌套)数据的变量关系研究,必须借助多层模型来实现。两层模型中,层一自变量Xij按组均值中心化,并将组均值X.j置于层2截距方程式中,可将Xij对因变量Yij的效应分解为组间和组内部分,二者之差被称为情境效应,X.j称为情境变量。多层结构方程模型(MSEM)将多层线性模型(MLM)和结构方程模型(SEM)相结合,通过设置潜变量和多指标的方法校正了MLM在情境效应分析中出现的抽样误差和测量误差,同时解决了数据的多层(嵌套)结构和潜变量的估计问题。除了分析原理的说明,还以班级平均竞争氛围对学生竞争表现的情境效应为例进行分析方法的示范,并比较MSEM和MLM的异同,随后展望了MSEM情境效应模型、情境效应无偏估计方法和情境变量研究的拓展方向。
The advantage of multilevel modeling （MLM） is to examine the data with nested structure which is frequently collected in the social and educational research. The effect of level-1 independent variable Xij on Yy would be decomposed into within-group and between-group effects by group mean centering with the means reintroduced into the level-2 intercept model, whose difference was called contextual effect and the reintroduced mean score of independent variable Xj was called contextual variable. In order to estimate the complex structure of sampling error and measurement error occurring in multilevel modeling, multilevel structural equation modeling （MSEM） treats the contextual variable as latent with multiple indicators. The superiority of MSEM is the integration of MLM and SEM to simultaneously analyze nested data structure and estimate latent variable. By taken as an example, the contextual effects of Competitive atmosphere of the class on Competitive behavior of individual students was introduced for illustrating how the analysis of contextual effects was achieved by MSEM. The future research direction of contextual effects, unbiased estimation of the effect and contextual variable is discussed at the end of the paper.
出处
《心理科学进展》
CSSCI
CSCD
北大核心
2011年第2期284-292,共9页
Advances in Psychological Science
基金
教育部人文社科基地项目(2009JJDXLX006)
广东省自然科学基金项目(9151063101000002)资助
关键词
多层线性模型
多层结构方程模型
情境效应
抽样误差
测量误差
multilevel model
multilevel structural equations modeling
contextual effect
contextual variable
sampling error
measurement error
