Date of Award


Document Type

Open Access Dissertation



First Advisor

Timothy Hanson


Reliability and survival data are widely encountered across many common settings. Subjects under investigation often include machines, bioassays, patients, etc.; their reliability or survival distribution, and its association with covariate processes, are commonly of interest. Within this dissertation, the first two chapters focus on reliability data where repairable systems fail and get interventions, e.g. repairs in the event process. It begins with a nonparametric test for the commonly assumed ''good as old'' assumption for minimal repair models and then a semi-parametric regression model is introduced for reliability data using Kijima's effective age. The third chapter focuses on survival data observed with potential spatial correlation. We first develop a Bayesian semi-parametric approach to the extended hazard model and then extend this framework to allow for spatial correlation among survival times. In contrast to widely used frailty models, our approach preserves marginal interpretations. Flexible modeling approaches in the Bayesian context are used for baseline failure rate or hazard and Markov chain Monte Carlo techniques to obtain the posterior inferences. The proposed tests and models are examined in several simulation studies and applications.