Dissertation/Thesis Abstract

Topics in model-based population inference
by Schutt, Rachel, Ph.D., Columbia University, 2010, 158; 3400625
Abstract (Summary)

This thesis comprises two parts: (I) estimating transmission rates on a network; and (II) poststratification as a unified framework for the analysis of sample surveys.

In Part I, we solve the problem of estimating person-to-person transmission rates of a contagious process on a network. We conduct a simulation study to demonstrate that despite the infection times of the individuals being missing data, we are able to estimate the transmission rates. We apply our method to a data set where the individuals in the network are babies in a neo-natal intensive care unit and the disease of interest is MRSA (Methicillin-resistant Staphycocculus Aurerus).

In Part II, we pose the problem of developing a Bayesian unified framework for the analysis of sample survey data that incorporates uncertainty about design and census numbers, and we formulate such a framework. Using this framework, it is now possible to estimate population means and population regression functions using poststratification even when census numbers are unavailable. Further, it is now possible to incorporate survey weights into a model-based analysis. We illustrate our methods on a data set on attitudes towards gay rights. We demonstrate that we can estimate the nonlinear pattern of support for gay unions, as a function of age of respondent, properly including the information in survey weights.

Indexing (document details)
Advisor: Gelman, Andrew, Dolgoarshinnykh, Regina
School: Columbia University
School Location: United States -- New York
Source: DAI-B 71/03, Dissertation Abstracts International
Subjects: Statistics, Epidemiology
Keywords: Population inference, Poststratification, Transmission rates
Publication Number: 3400625
ISBN: 9781109673111
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy