Approximate Bayesian approaches and semiparametric methods for handling missing data

Thumbnail Image
Date
2018-01-01
Authors
Sang, Hejian
Major Professor
Advisor
Jae Kwang Kim
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Organizational Unit
Statistics
As leaders in statistical research, collaboration, and education, the Department of Statistics at Iowa State University offers students an education like no other. We are committed to our mission of developing and applying statistical methods, and proud of our award-winning students and faculty.
Journal Issue
Is Version Of
Versions
Series
Department
Abstract

This thesis consists of four research papers focusing on estimation and inference in missing data. In the first paper (Chapter 2), an approximate Bayesian approach is developed to handle unit nonresponse with parametric model assumptions on the response probability, but without model assumptions for the outcome variable. The proposed Bayesian method is also extended to incorporate the auxiliary information from full sample. In second paper (Chapter 3), a new Bayesian method using the Spike-and-Slab prior is proposed to handle the sparse propensity score estimation. The proposed method is not based on any model assumption on the outcome variable and is computationally efficient. In third paper (Chapter 4), we develop a robust semiparametric method based on the profile likelihood obtained from semiparametric response model. The proposed method uses the observed regression model and the semiparametric response model to achieve robustness. An efficient algorithm using fractional imputation is developed. The bootstrap testing procedure is also proposed to test ignorability assumption. In last paper (Chapter 5), we propose a novel semiparametric fractional imputation method using Gaussian mixture model for handling multivariate missingness. The proposed method is computationally efficient and leads to robust estimation. The proposed method is further extended to incorporate the categorical auxiliary information. Asymptotic properties are developed for each proposed methods. Both simulation studies and real data applications are conducted to check the performance of the proposed methods in this thesis.

Comments
Description
Keywords
Citation
DOI
Source
Subject Categories
Copyright
Sun Apr 01 00:00:00 UTC 2018