Georgia Southern University
Survival Function Precision Under Censoring: A Comparative Study of Cox and Weibull Aft Models
Abstract
dc:description.abstractThis dissertation investigates the estimation and comparative performance of survival functions using the Cox Proportional Hazards (PH) model and the Weibull Accelerated Failure Time (AFT) model, two widely used frameworks in survival analysis. The study emphasizes quantile-based analysis to evaluate survival probabilities and their confidence intervals at key points across the time distribution. Through simulation experiments and application to the NCCTG lung cancer dataset, the research reveals that while both models yield comparable point estimates, the Weibull model consistently provides more stable and narrower confidence intervals, particularly in the tail regions of the survival curve. This is due to its parametric form, as it confers robustness when there is censoring and sparse data. The Cox model, though flexible with no assumption, has higher variance in later quantiles since it has non-parametric estimation of the baseline hazard. The findings contribute to evidence-based model selection in survival analysis and underscore the trade-off between flexibility and precision in parametric versus semi-parametric approaches.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Public Health in Biostatistics (Dr.P.H.)
- Level thesis:degree_level
- Dissertation (restricted to Georgia Southern)
- Discipline thesis:degree_discipline
- Department of Biostatistics, Epidemiology, and Environmental Health Sciences
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Alliu, Ibrahim L
- Contributors dc:contributor
-
- Hani Samawi
- Jing Kersey
Subjects
dc:subject × 6Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.georgiasouthern.edu/etd/2982