Bayesian Approach to Handling Informative Sampling

TL;DR

This paper presents a Bayesian framework for addressing informative sampling by integrating classical sampling theory with Bayesian analysis, enabling better population inference and model identification.

Contribution

It introduces a novel Bayesian approach that combines sampling theory and Bayesian methods to handle informative sampling and improve model estimation.

Findings

Proposed method effectively estimates population models under informative sampling.

Incorporates known inclusion probabilities into Bayesian modeling.

Shows promising results in model identification and estimation.

Abstract

In the case of informative sampling the sampling scheme explicitly or implicitly depends on the response variable. As a result, the sample distribution of response variable can- not be used for making inference about the population. In this research I investigate the problem of informative sampling from the Bayesian perspective. Application of the Bayesian approach permits solving the problems, which arise due to complexity of the models, being used for handling informative sampling. The main objective of the re- search is to combine the elements of the classical sampling theory and Bayesian analysis, for identifying and estimating the population model, and the model describing the sam- pling mechanism. Utilizing the fact that inclusion probabilities are generally known, the population sum of squares of the models residuals can be estimated, implementing the techniques of the sampling theory. In this research I show, how these estimates can be incorporated in the Bayesian modeling and how the Full Bayesian Significance Test (FBST), which is based on the Bayesian measure of evidence for precise null hypothesis, can be utilized as a model identification tool. The results obtained by implementation of the proposed approach to estimation and identification of the sample selection model seem promising. At this point I am working on methods of estimation and identification of the population model. An interesting extension of my approach is incorporation of known population characteristics into the estimation process. Some other directions for continuation of my research are highlighted in the sections which describe the proposed methodology.