東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Regression methods of time-dependent...

Hu, Nan.

Linked to FindBook

Google Book

Amazon

博客來

Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers.

Record Type:	Language materials, printed : Monograph/item
Title/Author:	Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers./
Author:	Hu, Nan.
Description:	240 p.
Notes:	Source: Dissertation Abstracts International, Volume: 71-05, Section: B, page: 3123.
Contained By:	Dissertation Abstracts International71-05B.
Subject:	Biology, Biostatistics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3406130
ISBN:	9781109727876

Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers.
Hu, Nan.

Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers. - 240 p.

Source: Dissertation Abstracts International, Volume: 71-05, Section: B, page: 3123.

Thesis (Ph.D.)--University of Washington, 2010.

Receiver operating characteristic (ROC) curves are commonly used for visualizing sensitivity and specificity of a continuous biomarker or diagnostic test result, Y, for a binary disease outcome D. In practice, however, many disease outcomes depend on time. Therefore, it is appropriate to derive the corresponding time-dependent ROC curves. In this work. I first introduce a new semi-parametric regression approach for estimating the covariate adjusted time-dependent ROC curves by modeling time-dependent sensitivities, or true positive rates (TPRs), and time-dependent false positive rates (FPRs), based on a transformation model for the event time, T, and a semi-parametric location model for the biomarker, Y. I further discuss the new method according to whether the disease time, T, is subject to censoring. Different transformation model is used for the two situations. Since the transformation model does not place any assumptions on the distribution of an event time outcome, this approach can be applied to more general case and is more robust than previous semi-parametric methods. Numerical study was implemented for the heteroscedastic transformation model when the error term follows the standard extreme value distribution, the standard normal distribution and the logistic distribution. The results show that our estimator is unbiased and robust to mis-specification of the time-to-event model. The efficiency is comparable with the correctly specified model and much higher than the mis-specified model. The new method was applied to analyze data from HIVNET 012 randomized trial for evaluating the two biomarkers of predicting mother-to-infant transmission of HIV-1 virus, and to analyze data front VA lung cancer trial for evaluating the performance score of predicting the lung cancer event. The regression approach for censored disease time was applied to VA Lung Cancer Trial to evaluate biomarkers for predicting the mortality of the study subjects. The other regression approach I proposed is a directly modeling method for the time-dependent sensitivity (ROC curve) at a given specificity for biomarkers with repeated measurements. I show, in this work, that the direct time-dependent ROC model is equivalent to a transformation model with unknown transformation function and error distribution. The proposed semi-parametric ROC model have a good interpretation for its regression parameters and is relatively easy to implement. Numerical studies showed that the proposed estimator is unbiased when the biomakers data are completely balanced and is missing completely at random in a monotone pattern. The proposed ROC model and estimation procedure is demonstrated using VAX004 HIV-1 vaccine trial.

ISBN: 9781109727876Subjects--Topical Terms:

1018416
Biology, Biostatistics.

Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers.
LDR:03655nam 2200289 4500 001 1402447
005 20111102135954.5
008 130515s2010 ||||||||||||||||| ||eng d
020 $a 9781109727876
035 $a (UMI)AAI3406130
035 $a AAI3406130
040 $a UMI $c UMI
100 1 $a Hu, Nan. $3 1681637
245 1 0 $a Regression methods of time-dependent ROC curve for evaluating the prognosis capacity of biomarkers.
300 $a 240 p.
500 $a Source: Dissertation Abstracts International, Volume: 71-05, Section: B, page: 3123.
500 $a Adviser: Xiao-Hua Zhou.
502 $a Thesis (Ph.D.)--University of Washington, 2010.
520 $a Receiver operating characteristic (ROC) curves are commonly used for visualizing sensitivity and specificity of a continuous biomarker or diagnostic test result, Y, for a binary disease outcome D. In practice, however, many disease outcomes depend on time. Therefore, it is appropriate to derive the corresponding time-dependent ROC curves. In this work. I first introduce a new semi-parametric regression approach for estimating the covariate adjusted time-dependent ROC curves by modeling time-dependent sensitivities, or true positive rates (TPRs), and time-dependent false positive rates (FPRs), based on a transformation model for the event time, T, and a semi-parametric location model for the biomarker, Y. I further discuss the new method according to whether the disease time, T, is subject to censoring. Different transformation model is used for the two situations. Since the transformation model does not place any assumptions on the distribution of an event time outcome, this approach can be applied to more general case and is more robust than previous semi-parametric methods. Numerical study was implemented for the heteroscedastic transformation model when the error term follows the standard extreme value distribution, the standard normal distribution and the logistic distribution. The results show that our estimator is unbiased and robust to mis-specification of the time-to-event model. The efficiency is comparable with the correctly specified model and much higher than the mis-specified model. The new method was applied to analyze data from HIVNET 012 randomized trial for evaluating the two biomarkers of predicting mother-to-infant transmission of HIV-1 virus, and to analyze data front VA lung cancer trial for evaluating the performance score of predicting the lung cancer event. The regression approach for censored disease time was applied to VA Lung Cancer Trial to evaluate biomarkers for predicting the mortality of the study subjects. The other regression approach I proposed is a directly modeling method for the time-dependent sensitivity (ROC curve) at a given specificity for biomarkers with repeated measurements. I show, in this work, that the direct time-dependent ROC model is equivalent to a transformation model with unknown transformation function and error distribution. The proposed semi-parametric ROC model have a good interpretation for its regression parameters and is relatively easy to implement. Numerical studies showed that the proposed estimator is unbiased when the biomakers data are completely balanced and is missing completely at random in a monotone pattern. The proposed ROC model and estimation procedure is demonstrated using VAX004 HIV-1 vaccine trial.
590 $a School code: 0250.
650 4 $a Biology, Biostatistics. $3 1018416
650 4 $a Statistics. $3 517247
650 4 $a Health Sciences, Public Health. $3 1017659
690 $a 0308
690 $a 0463
690 $a 0573
710 2 $a University of Washington. $3 545923
773 0 $t Dissertation Abstracts International $g 71-05B.
790 1 0 $a Zhou, Xiao-Hua, $e advisor
790 $a 0250
791 $a Ph.D.
792 $a 2010
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3406130