Multiple Regression Analysis with Data from Complex Survey

• Activity Number: 664
• Type: Contributed
• Date/Time: Thursday, August 4, 2011 : 10:30 AM to 12:20 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Esher Hsu*+
• Companies: National Taipei University
• Keywords: Multiple Regression Analysis ; Stratified weighted least squares estimator ; Probability weighted least squares estimator ; Quasi-Aitken weighted least square estimator ; Complex Survey ; Social Change

Abstract

This study explores multiple regression analysis with complex survey data. Four methods of multiple regression analysis, namely, ordinary least squares, weighted least squares, probability weighted least squares, and Quasi-Aitken probability weighted least squares are proposed for comparison by Monte Carlo approach to compare their efficiency based upon bias, variance, and MSE. The data from "Taiwan Social Change Survey 2007" collected under a stratified unequal probability sampling were used for empirical analysis to compare four proposed methods based upon the estimated regression coefficients and RMSE. The simulation results show that probability weighted least squares estimator and Quasi-Aitken weighted least square estimator perform better than others under the unequal probability design. The empirical results consist with the simulation results. The empirical results show that the education years of respondents in Taiwan has significant negative relationship with their age but has positive relationship with their parents' education years.