東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Variable selection in the general li...

Yu, Lili.

FindBook

Google Book

Amazon

博客來

Variable selection in the general linear model for censored data.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Variable selection in the general linear model for censored data./
作者:	Yu, Lili.
面頁冊數:	140 p.
附註:	Adviser: Dennis K. Pearl.
Contained By:	Dissertation Abstracts International68-01B.
標題:	Statistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3247960

Variable selection in the general linear model for censored data.
Yu, Lili.

Variable selection in the general linear model for censored data. - 140 p.

Adviser: Dennis K. Pearl.

Thesis (Ph.D.)--The Ohio State University, 2007.

Variable selection is a popular topic in statistics today. However, for right censored data, only a few methods are available. The principle method assumes that the data comes from a Cox proportional hazards model. In 1997, Tibshirani proposed a variation of the LASSO method that minimizes the log partial likelihood subject to the sum of the absolute values of the parameters being bounded by a constant in the Cox proportional hazards model. Due to the nature of this constraint, it shrinks coefficients and produces some coefficients that are exactly zero. The resulting prediction error is smaller than that of subset selection methods. However, the proportional hazard assumption isn't always appropriate for real data. Therefore, we apply this method to the class of models (linear regression models) in which the response variable is right censored and the error is symmetric at zero, but is otherwise distribution free. The method also uses a sieve-likelihood to calculate a variation of the LASSO criterion and uses generalized cross-validation to choose the tuning parameter. Simulation shows that this method gives smaller prediction error than the method that depends on the proportional hazard assumption in some scenarios, especially for larger sample sizes. The performance of the proposed method is also examined via a data set from a study of the ganglioside content of primary brain tumors and a data set from a study of bone marrow transplants in Chronic Myelogenous Leukemia patients.Subjects--Topical Terms:

517247
Statistics.

Variable selection in the general linear model for censored data.
LDR:02326nam 2200253 a 45 001 963716
005 20110831
008 110831s2007 eng d
035 $a (UMI)AAI3247960
035 $a AAI3247960
040 $a UMI $c UMI
100 1 $a Yu, Lili. $3 1286779
245 1 0 $a Variable selection in the general linear model for censored data.
300 $a 140 p.
500 $a Adviser: Dennis K. Pearl.
500 $a Source: Dissertation Abstracts International, Volume: 68-01, Section: B, page: 0374.
502 $a Thesis (Ph.D.)--The Ohio State University, 2007.
520 $a Variable selection is a popular topic in statistics today. However, for right censored data, only a few methods are available. The principle method assumes that the data comes from a Cox proportional hazards model. In 1997, Tibshirani proposed a variation of the LASSO method that minimizes the log partial likelihood subject to the sum of the absolute values of the parameters being bounded by a constant in the Cox proportional hazards model. Due to the nature of this constraint, it shrinks coefficients and produces some coefficients that are exactly zero. The resulting prediction error is smaller than that of subset selection methods. However, the proportional hazard assumption isn't always appropriate for real data. Therefore, we apply this method to the class of models (linear regression models) in which the response variable is right censored and the error is symmetric at zero, but is otherwise distribution free. The method also uses a sieve-likelihood to calculate a variation of the LASSO criterion and uses generalized cross-validation to choose the tuning parameter. Simulation shows that this method gives smaller prediction error than the method that depends on the proportional hazard assumption in some scenarios, especially for larger sample sizes. The performance of the proposed method is also examined via a data set from a study of the ganglioside content of primary brain tumors and a data set from a study of bone marrow transplants in Chronic Myelogenous Leukemia patients.
590 $a School code: 0168.
650 4 $a Statistics. $3 517247
690 $a 0463
710 2 0 $a The Ohio State University. $3 718944
773 0 $t Dissertation Abstracts International $g 68-01B.
790 $a 0168
790 1 0 $a Pearl, Dennis K., $e advisor
791 $a Ph.D.
792 $a 2007
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3247960