Performance of First-Order Weighted Random Effects Model Estimators Under Complex Sampling Designs

• Activity Number: 343
• Type: Contributed
• Date/Time: Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Yue Jia*+ and S. Lynne Stokes and Ian Harris
• Companies: Educational Testing Service and Southern Methodist University and Southern Methodist University
• Keywords: weighting ; one-way random effects model ; estimation bias ; sample survey ; NAEP

Abstract

We consider estimation of parameters of random effects models from a sample collected via a complex multi-stage design. We evaluate the first-order weighted method of moments estimators for estimating the grand mean and two variance components. For these estimators, analytic bias expressions are developed and the accuracy of the expressions is evaluated through Monte Carlo simulation. These expressions are then used to examine the impact of sample size, ICC and the sampling design effects. Our study shows that the first-order weighting method effectively reduces the bias for the estimator of mean, and performs well to estimate the variance components for moderate sample size and ICC level. However, for the variance components estimation, this weighting method should be used with caution for small cluster sizes, particularly with low ICC levels.