|
Activity Number:
|
474
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301306 |
|
Title:
|
Multiple Imputation for Latent Variables in Classical Test Theory for Cluster Sample
|
|
Author(s):
|
Tiandong Li*+ and Robert Mislevy+
|
|
Companies:
|
Westat, Inc. and University of Maryland, College Park
|
|
Address:
|
1650 Research Blvd., Rockville, MD, 20850, 1230C Benjamin Building , College Park, MD, 20742,
|
|
Keywords:
|
Classical test theory ; Cluster sample ; Multiple imputation ; latent variable
|
|
Abstract:
|
Mislevy's 1991 Psychometrica paper introduced a model in making inferences about the distribution of a latent construct ? under complex sample design by applying "multiple imputation" techniques to Large Scale Assessment, such as NAEP. That presentation developed the case for covariates treated as fixed effects, and mentioned the extension to random effects in passing. This study extends the discussion of random effects under the classical test theory framework. We developed analytical solution for the mean and variance of the imputation on for clustered samples, following Rubin's formula for Multiple Imputation. We explored the effect on the variance of the mean of the imputed values by changing the relative sizes of the three variance components, between cluster variance, within cluster variance and error variance, and sample size. Findings were also confirmed by a simulation study.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2008 program |
