Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


Tao_dissertation.pdf (4.25 MB)

ETD Abstract Container

Abstract Header

Bayesian Threshold Regression for Current Status Data with Informative Censoring

Xiao, Tao
http://rave.ohiolink.edu/etdc/view?acc_num=osu1438272888

Abstract Details

2015, Doctor of Philosophy, Ohio State University, Public Health.
In some biomedical studies, there is interest in making inferences about a time to event distribution but the exact time of the event is unknown. For instance in carcinogenicity studies in animals, tumors are not discovered until the time of examination and hence time to tumor is left censored; this is known as current status data. Sometimes, the examination time is not independent of the event time. For example, in an animal study, an exam may have occurred because the animal died due to a cause related to tumor development. In this case, survival analysis methods which assume independent censoring would result in biased inferences. To address this issue, we propose a Bayesian approach which jointly models time to event and time to censoring using latent Wiener processes which fail once they hit a boundary value. Using data augmentation, we sample the unobserved event time and values of the latent processes for those subjects who do not experience an event. Informative censoring is accounted for also by modeling time to censoring using latent health process. Sometimes multivariate current status data also arise, e.g., tumors can develop in multiple organ sites in carcinogenicity studies. Examination time occurring at natural death could be affected by these different types of tumors which may intrinsically correlate with each other. We propose a multivariate Bayesian approach to accommodate multiple left censored events driven by different latent Wiener processes. We use a random effect shared by the drifts of the processes underlying the events of interest to model the correlation of the event times. The censoring process is modeled using a latent Wiener process whose time scale is affected by the occurrence of an event thus accounting for dependent censoring. Our models are conceptually appealing and do not require the assumption of proportional hazards of some standard methods. In simulation studies, we found that the proposed informative censoring models provide more accurate estimates of regression coefficients than the independent censoring models when the data do not satisfy the assumption of independent censoring. We applied our methods to data from National Toxicology Program studies.
Michael Pennell (Advisor)
Grzegorz Rempala (Committee Member)
Soledad Fernandez (Committee Member)
183 p.
Biostatistics; Public Health

Recommended Citations

Citations

  • Xiao, T. (2015). Bayesian Threshold Regression for Current Status Data with Informative Censoring [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1438272888

    APA Style (7th edition)

  • Xiao, Tao. Bayesian Threshold Regression for Current Status Data with Informative Censoring. 2015. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1438272888.

    MLA Style (8th edition)

  • Xiao, Tao. "Bayesian Threshold Regression for Current Status Data with Informative Censoring." Doctoral dissertation, Ohio State University, 2015. http://rave.ohiolink.edu/etdc/view?acc_num=osu1438272888

    Chicago Manual of Style (17th edition)

osu1438272888
861
© 2015, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.