Degree Type

Dissertation

Date of Award

2012

Degree Name

Doctor of Philosophy

Department

Statistics

First Advisor

Alicia Carriquiry

Second Advisor

Chong Wang

Abstract

The thesis is composed of three separated projects: disease risk scoring systems (chapter 2), statistical tests for proportion difference in one-to-two matched binary data (chapter 3) and bivariate measurement error model for nutrition epidemiology (chapter 4). In the first project, we propose to use group lasso algorithm for logistic regression to construct a risk scoring system for predicting disease in swine. We choose the penalty parameter for the group lasso through leave-one-out cross validation and use the area under the receiver operating characteristic curve as criterion. We show our proposed scoring system is superior to existing methods. The second project was originally motivated by the pooling of diagnostic tests. We proposed exact and asymptotic tests for one-to-two matched binary data. Unlike other existing methods, our procedure doesn't rely on a mutual independence assumption. The emphasis on dependence among observations from the same matched set is natural and appealing, as much in human health as it is in veterinary medicine. It can be applied to many kinds of diagnostic studies with a one-to-two matched data structure. Our method can also be generalized to one-to-N matched case in a straightforward manner. In the third paper we consider the problem of estimating the joint distribution of two correlated random variables where one of the variables is observed with error. DKM is first used to adjust the univariate measurement error. A Gaussian copula is then used to model the correlation structure between the two variables after error adjustment.

DOI

https://doi.org/10.31274/etd-180810-16

Copyright Owner

Hui Lin

Language

en

File Format

application/pdf

File Size

78 pages

Share

COinS