WITHDRAWN: Robust Bayesian Approach to the Analysis of Finite Population Surveys

• Author(s): Jairo Fuquene
• Keywords: Survey Sampling ; Robust Priors ; Horvitz-Thompson Estimator

Abstract

The proposal in this work is to use Cauchy and Berger's robust priors in the Bayesian models to the survey designs. The robust Bayesian Approach is more effective than the particular case of noninformative priors, as suggested Little and Zheng (2007), because when there is a conflict between prior information and the sample information the Cauchy and Berger's priors effectively becomes a noninformative prior and in this precise sense the prior information is discounted. The behavior of Robust Bayesian methods is qualitative different than noninformative Bayesian methods and arguably much more reasonable and acceptable to the people involved in survey sampling. Finally, we apply this methodology in real and MCMC simulated examples.