|
Activity Number:
|
467
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 9, 2006 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Social Statistics Section
|
| Abstract - #307024 |
|
Title:
|
Multilevel Structural Equation Model for Ordinal Responses
|
|
Author(s):
|
Sophia Rabe-Hesketh*+ and Xiaohui Zheng
|
|
Companies:
|
University of California, Berkeley
|
|
Address:
|
3659 Tolman Hall, Berkeley, CA, 94720,
|
|
Keywords:
|
multilevel model ; mixed model ; item response model ; structural equation model ; latent variable ; GLLAMM
|
|
Abstract:
|
In a standard multilevel model for subjects nested in clusters, a single response variable is regressed on subject-level and cluster-level covariates and random effects represent any between-cluster heterogeneity not accommodated by the covariates. Here we extend these models to allow the response and/or covariates to be latent variables that are measured indirectly by several ordinal indicators. Item-response models for the relationship between indicators and latent variables are combined with structural models relating the latent response to the latent and observed covariates. The models are applied to the U.S. sample of the Program for International Student Assessment 2000 to investigate the relationship between the student-level latent variable 'reading ability' and the school-level latent variable 'teacher excellence', controlling for a number of observed student-level covariates.
|
