Date of Award

1-1-2012

Document Type

Campus Access Thesis

Department

Epidemiology and Biostatistics

Sub-Department

Biostatistics

First Advisor

Hongmei Zhang

Abstract

To detect association between Single Nucleotide Polymorphism (SNP) and disease, extensive investigations have been conducted. Most work has focused on testing SNP effect one by one due to the difficulty of including a large number of SNPs in a statistical model. In this thesis, we explored a semi-parametric method built upon kernel machines to include multiple SNPs in a model. The information of SNPs similarity calculated based on identity-by-state (IBS) algorithm was included in a kernel matrix. We evaluated the method via various scenarios based on simulation studies. The method is efficient to estimate and test SNP effects. The testing power increases with the increase of sample size and the strength of SNP effect. The testing power also increases with the increase of SNPs similarity level and decrease of sparsity level of kernel matrix. We applied this method to a SNPs-Lung function data to test the SNP effects on lung function measured by Forced Vital Capacity.

Share

COinS