|
Activity Number:
|
492
|
|
Type:
|
Invited
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #302830 |
|
Title:
|
Borrowing Strength Over Space in Small Area Estimation: Comparing Local and Spline-Based M-Quantile Models with Spatial Autoregressive Random Effects Models
|
|
Author(s):
|
Nikos Tzavidis*+
|
|
Companies:
|
University of Manchester
|
|
Address:
|
Kantorovich Building, Manchester, International, M13 9PL, United Kingdom
|
|
Keywords:
|
Spatial correlation ; Robust estimation ; Local smoothing ; Small area prediction ; Linear mixed model ; Random effects
|
|
Abstract:
|
A popular approach to incorporating spatial information in small area estimation models between area variability via a Simultaneous Autoregressive (SAR) random effects model. An alternative method that incorporates spatial information locally is M-quantile Geographically Weighted Regression (GWR), which fits a local model to the regression M-quantiles of the conditional distribution of the outcome variable. A more global approach uses spline approximations to fit nonparametric M-quantile regression models that reflect spatial variation in the data. In this presentation we contrast SAR, M-quantile GWR and M-quantile spline models in terms of their performance using data with different levels of geographical detail. Our analysis is illustrated using data from the Environmental Monitoring and Assessment Program of the EPA and the World Bank's Albanian Living and Standards Measurement Study.
|
