|
Activity Number:
|
197
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 1, 2016 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Survey Research Methods Section
|
|
Abstract #321024
|
|
|
Title:
|
Comparison of MCMC and ADM Methods for Hierarchical Bayesian Estimates in Small-Area Estimation
|
|
Author(s):
|
Ying Han*
|
|
Companies:
|
University of Maryland
|
|
Keywords:
|
MCMC ;
ADM ;
Hierarchical ;
Gibbs Sampling ;
Small Area Estimation
|
|
Abstract:
|
We will compare the accuracy of the Monte Carlo Markov Chain (MCMC) and Adjusted Density Method (ADM) in approximating Hierarchical Bayesian (HB) solution in the context of small area estimation. For this comparison, we will treat the numerical integration method as the golden standard. We apply the hierarchical model to the poverty data from the 2005 American Community Survey. We use both MCMC and ADM to obtain approximated HB estimates of the poverty rate for the 0-17 year old children in each county. We also experiment with different priors. Both MCMC and ADM methods can be used to approximate the posterior distributions. MCMC uses the Gibbs Sampling Algorithm and ADM approximates a posterior density with a Pearson density. Plots of the MCMC and ADM approximations to the posterior distributions are shown respectively. Both MCMC and AMD methods provide strikingly similar results. The ADM method is very fast compared to the MCMC method and should be appealing when several posterior distributions need to be computed in a simulation environment. This work is completed under supervision of Professor Partha Lahiri.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2016 program
|
