School Seminars and Colloquia

A Cox-log-linear model for competing risk data

Statistics Seminar

by Dr Meei Ng


Institution: Department of Mathematics and Statistics, The University of Melbourne
Date: Thu 24th May 2007
Time: 12:00 PM
Location: Room 213, Richard Berry Building

Abstract: Competing risk data arise frequently in medical studies where failures can be classified into various types. The current methods of analyzing such data are based on modeling each cause-specific hazard function on its own using standard survival analysis techniques, or non-parametric pairwise comparison of the cumulative incidence functions of treatment groups. These methods have disadvantages that are well known, and will be discussed. We propose a new method of modelling that provides a single baseline hazard function, and allows parametric overall (and pairwise) comparison of cause-specific hazard functions and cumulative incidence functions. It also enables the incorporation of covariate information. An example will be used to illustrate the procedures.

For More Information: Dr Owen Jones: odj@mailhost.ms.unimelb.edu.au