|
Activity Number:
|
284
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #304072 |
|
Title:
|
Model-Based Methods in Analyzing Complex Survey Data: A Case Study with National Health Interview Survey Data
|
|
Author(s):
|
Rong Wei*+ and Van L. Parsons
|
|
Companies:
|
National Center for Health Statistics and National Center for Health Statistics
|
|
Address:
|
3311 Toledo Rd, Room 3114 , Hyattsville, MD, 20782,
|
|
Keywords:
|
general linear mixed models ; multilevel modeling ; design-based
|
|
Abstract:
|
In this study we consider model-based methods that can be used to account for clustering, stratification and weighting effects in complex-survey-design data. Generalized linear mixed effect models were developed based on the adult sample from the public-release of 2007 National Health Interview Survey (NHIS). For this public-release there are only two available levels of clustering, strata and Primary Sampling Units (PSUs) for use in the model-based method. Model-based multilevel variance/covariance structures were estimated using algorithms given in the SAS procedure GLIMMIX. These model-based methods will be compared empirically with the design-based method of the SUDAAN software, as well as with a fixed effect model in the SAS procedure LOGISTIC.
|
