Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

HIGH DIMENSIONAL VARIABLE SELECTION VIA PENALIZED LIKELIHOOD FOR GENERALIZED LINEAR MODELS

Qi, Wenjing (2015) HIGH DIMENSIONAL VARIABLE SELECTION VIA PENALIZED LIKELIHOOD FOR GENERALIZED LINEAR MODELS. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img] PDF
Primary Text
Restricted to University of Pittsburgh users only until 14 January 2020.

Download (1MB) | Request a Copy

Abstract

Variable selection is fundamental to high dimensional statistical modeling. In this study, penalized likelihood methods are examined to simultaneously estimate parameters and select variables for generalized linear models. We focus on the variable selection and parameter estimation properties rather than the prediction properties of the estimators and are more interested in situations where the number of parameters diverges with the sample size. We prove the parameter estimation consistency of several widely used penalized likelihood estimators for generalized linear models. We define the relaxed sense and prove that it loosens the regularity and sparsity conditions of the parameter estimation and variable selection consistency. We propose a bootstrap method that can greatly improve the variable selection performances and reduce false discovery rates. We conduct simulation studies to compare the variable selection and parameter estimation properties of these penalized likelihood estimators for logistic models. We then illustrate our methods on gene expression data.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Qi, Wenjingwq4@pitt.eduWQ4
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairIyengar, Satishssi@pitt.eduSSI
Committee MemberGleser, Leon J.gleser@pitt.eduGLESER
Committee MemberCheng, Yuyucheng@pitt.eduYUCHENG
Committee MemberAnderson, Stewart J.sja@pitt.eduSJA
Date: 14 January 2015
Date Type: Publication
Defense Date: 25 August 2014
Approval Date: 14 January 2015
Submission Date: 3 December 2014
Access Restriction: 5 year -- Restrict access to University of Pittsburgh for a period of 5 years.
Number of Pages: 77
Institution: University of Pittsburgh
Schools and Programs: Dietrich School of Arts and Sciences > Statistics
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Variable selection, High dimensionality, Penalized methods, Generalized linear models, Bootstrap method.
Date Deposited: 14 Jan 2015 15:38
Last Modified: 15 Nov 2016 14:25
URI: http://d-scholarship.pitt.edu/id/eprint/23758

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item