|
Activity Number:
|
693
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 4, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
|
Abstract #318993
|
View Presentation
|
|
Title:
|
Multilevel Quantile Regression Models for Complex Surveys
|
|
Author(s):
|
Jing Wang* and John Fu
|
|
Companies:
|
St. Louis University and Saint Louis University
|
|
Keywords:
|
Multilevel model ;
Sampling weight ;
Gibbs sampling ;
Conditional quantile ;
Asymmetric Laplace density ;
Bootstrapping
|
|
Abstract:
|
Multilevel modeling has been applied to data from complex surveys using design effects including probability sampling weights, stratification and clustering. We apply this approach to the conditional quantile of a continuous outcome from the complex survey data in the Bayesian framework. Gibbs sampling is implemented to derive posterior likelihood functions based on the asymmetric Laplace density. Maximization of the posterior likelihood is accomplished by using the Nelder-Mead simplex method that accounts for design effects. Bootstrapping is used to obtain standard errors of the quantile regression coefficient estimates. This approach is illustrated using data on childhood BMI from the 2011-2012 National Health and Nutrition Examination Survey (NHANES) and a Monte Carlo simulation study. Our results show that the weighted Bayesian QR estimator provides a more comprehensive picture of the effects of the covariates on the distribution of the response than the weighted mean regression estimator for complex surveys.
|
Authors who are presenting talks have a * after their name.
