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

Estimating causal treatment effects ...

The Johns Hopkins University.

FindBook

Google Book

Amazon

博客來

Estimating causal treatment effects for post-randomization marker data with failure event censoring.

紀錄類型:	書目-語言資料,印刷品 : Monograph/item
正題名/作者:	Estimating causal treatment effects for post-randomization marker data with failure event censoring./
作者:	Ning, Jing.
面頁冊數:	137 p.
附註:	Adviser: Mei-Cheng Wang.
Contained By:	Dissertation Abstracts International68-11B.
標題:	Biology, Biostatistics. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoeng/servlet/advanced?query=3288509
ISBN:	9780549312208

Estimating causal treatment effects for post-randomization marker data with failure event censoring.
Ning, Jing.

Estimating causal treatment effects for post-randomization marker data with failure event censoring. - 137 p.

Adviser: Mei-Cheng Wang.

Thesis (Ph.D.)--The Johns Hopkins University, 2008.

Bivariate recurrent event data arise in various forms such as familial studies involving repeated events, or biomedical studies concerning two types of recurrent events. We propose semi-parametric joint models through latent variables and develop the estimating procedure for cumulative rate functions and regression parameters. In the proposed models, the dependence between two processes is characterized by latent variables and the censoring time is allowed to be associated with bivariate recurrent events through the latent variables. The major advantage of such models is that the distributions of both latent variables and the censoring are treated as nuisance parameters in the modelling and estimating procedure. The dependence between two processes is also of interest except for the features of marginal processes. Due to the complex structure of bivariate recurrent event data, standard dependence measures for bivariate events are subject to a serious limitation: loss of information. Accordingly, one new measure of local dependence, rate ratio, is defined with a simple conditional probability interpretation. Under the joint latent variable models, we clearly illustrate the estimating procedure for the rate ratio. Asymptotic properties of the proposed estimators are established. Simulation results demonstrate that the estimators perform well with practical sample sizes. Juvenile behavior data and CPCRA ddI/ddC data are used to illustrate the proposed methodology.

ISBN: 9780549312208Subjects--Topical Terms:

1018416
Biology, Biostatistics.

Estimating causal treatment effects for post-randomization marker data with failure event censoring.
LDR:04920nam 2200289 a 45 001 856582
005 20100709
008 100709s2008 ||||||||||||||||| ||eng d
020 $a 9780549312208
035 $a (UMI)AAI3288509
035 $a AAI3288509
040 $a UMI $c UMI
100 1 $a Ning, Jing. $3 1023406
245 1 0 $a Estimating causal treatment effects for post-randomization marker data with failure event censoring.
300 $a 137 p.
500 $a Adviser: Mei-Cheng Wang.
500 $a Source: Dissertation Abstracts International, Volume: 68-11, Section: B, page: 7050.
502 $a Thesis (Ph.D.)--The Johns Hopkins University, 2008.
520 $a Bivariate recurrent event data arise in various forms such as familial studies involving repeated events, or biomedical studies concerning two types of recurrent events. We propose semi-parametric joint models through latent variables and develop the estimating procedure for cumulative rate functions and regression parameters. In the proposed models, the dependence between two processes is characterized by latent variables and the censoring time is allowed to be associated with bivariate recurrent events through the latent variables. The major advantage of such models is that the distributions of both latent variables and the censoring are treated as nuisance parameters in the modelling and estimating procedure. The dependence between two processes is also of interest except for the features of marginal processes. Due to the complex structure of bivariate recurrent event data, standard dependence measures for bivariate events are subject to a serious limitation: loss of information. Accordingly, one new measure of local dependence, rate ratio, is defined with a simple conditional probability interpretation. Under the joint latent variable models, we clearly illustrate the estimating procedure for the rate ratio. Asymptotic properties of the proposed estimators are established. Simulation results demonstrate that the estimators perform well with practical sample sizes. Juvenile behavior data and CPCRA ddI/ddC data are used to illustrate the proposed methodology.
520 $a In randomization studies, treatment effect on a mark variable, a measurement evaluated at the time of a failure event, is an important index for evaluating the treatment efficacy. A difficulty in analyzing this type of data is that the values of the mark variable are not ob servable when the failure events are censored. Also, conditioning on the occurrence of the failure event which occurs after treatment, the mark variable is typically subject to selection bias. Thus, in general, comparisons based on mark variable measured at the post-treatment event do not have causal interpretation. Furthermore, when a failure time censoring is present, the marginal distribution of the mark variable may not be fully identifiable. This non-identifiability problem makes evaluating causal treatment effect even more difficult. In the second part of the dissertation we consider models and required assumptions for nonparametric estimation of causal treatment effects on mark variables. We develop analytical procedures by borrowing information from failure time data to correct selection bias. Formulating the problem by the principal stratification (Frangakis and Rubin, 2002) framework of causal inference, we verify that proposed treatment effects are principal causal effects. Also, the causal effect based on a conditional distribution of the mark variable, rather than the marginal distribution, is identifiable. Asymptotic properties of the proposed estimators are established. CPCRA ddI/ddC data are used to illustrate the proposed methodology.
520 $a The third part of the dissertation focuses on the evaluation of treatment effect on a mark process in the presence of an informative terminal event in two-arm randomized trials. The mark process data arise naturally from studies of disease progression, in which mark vari ables are measured repeatedly over patients' follow-up periods. In such studies, the data are often complicated by irregular measurement times and possible informative terminal events. Accordingly the treatment effect estimated by the standard survivors' mean models generally does not have causal interpretation due to the association between the mark process and terminal event. A causal survivors' mean model therefore is proposed to capture a causal effect on the mark process by borrowing information from terminal event data. Simulation studies are conducted to evaluate the performance of the developed model and method. An illustration of the method using CPCRA ddl/ddC data is provided.
590 $a School code: 0098.
650 4 $a Biology, Biostatistics. $3 1018416
690 $a 0308
710 2 $a The Johns Hopkins University. $3 1017431
773 0 $t Dissertation Abstracts International $g 68-11B.
790 $a 0098
790 1 0 $a Wang, Mei-Cheng, $e advisor
791 $a Ph.D.
792 $a 2008
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoeng/servlet/advanced?query=3288509