|
Activity Number:
|
502
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #307131 |
|
Title:
|
Use Sampling Weights in Hierarchical Modeling
|
|
Author(s):
|
Yue Jia*+ and S. Lynne Stokes
|
|
Companies:
|
Southern Methodist University and Southern Methodist University
|
|
Address:
|
Department of Statistical Science, 144 Heroy Science Hall, Dallas, TX, 75275-0332,
|
|
Keywords:
|
sampling weights ; hierarchical models ; NAEP ; bias
|
|
Abstract:
|
Large-scale survey data often arise from complex multi-stage designs with known but unequal selection probabilities at each level. Hierarchical model (HM) is widely used to analyze the correlation structure induced by cluster designs. In addition, Sampling weights, the inverse of the selection probabilities, are used to produce unbiased parameter estimators. This paper is concerned with the question of incorporating sampling weights in HM, with the objective of predicting when ignoring the weights will bias results. We develop expressions for the bias of estimators from one-way random effects model. Our study shows that the bias is related to not only the sample size and the population size, but also the covariance between sampling weights and some functions of the random effects. The study is applied to National Assessment of Educational Progress (NAEP) 2003 4th Grade Reading data.
|
