Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Methods for the analysis of censored data with a focus on comparisons arising in HIV studies

McKinnon, Elizabeth J. (2000) Methods for the analysis of censored data with a focus on comparisons arising in HIV studies. PhD thesis, Murdoch University.

PDF - Whole Thesis
Available Upon Request


Censored data methods play an integral role in the analysis of survival data. Although there exist many well-known techniques that are readily applicable to estimation and comparison of survival curves with standard right-censored data, there are fewer methods available for analysis of non-standard censored data. The methodologies developed in this thesis are largely motivated by several types of non-standard analyses undertaken as part of the Western Australian Human Immunodeficiency Virus (WAHIV) Cohort study.

The main body of the thesis begins with the development of a general framework for comparing functionals of survival curve estimates, with the comparisons based on location-type features of the curves. Derived using distribution-free estimating equations, the resultant tests for equality of survival provide good power against a variety of alternatives. Moreover, the techniques are sufficiently flexible to enable comparisons of survival using non-standard estimators. We thus utilize the approach in two subsequent applications, which focus on the assessment of therapy effectiveness.

The first of these applications is an analysis that attempts to reduce potential biases due to the observational nature of the data. This is achieved by means of retrospective modeling of HIV disease markers for identification of appropriate controls. Comparisons between cases and matched controls are then undertaken using composite weighted survival functions.

In the second application we consider the estimation and comparison of profiles of blood viral load detectability following initiation of potent antiretroviral treatment regimens. The profiles are obtained from a convolution of two distributions, that of the time it takes to become undetectable and that of the subsequent time it takes to return to a state of detectability. Aspects incorporated include the interval-censored nature of the data as a result of irregular visit times and the effects of covariates, including the sojourn time in the detectable state. An approach based on a modified likelihood is shown to have a number of desirable properties and can be readily fitted via a generalized linear model formulation.

Item Type: Thesis (PhD)
Murdoch Affiliation(s): Division of Science and Engineering
Notes: Note to the author: If you would like to make your thesis openly available on Murdoch University Library's Research Repository, please contact: Thank you.
Supervisor(s): James, Ian
Item Control Page Item Control Page