Paul Meier's Anatomy of the Cox Regression Model

• Activity Number: 550
• Type: Invited
• Date/Time: Wednesday, August 1, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Memorial
• Author(s): Theodore Karrison*+
• Companies: The University of Chicago
• Keywords: Cox regression model ; Hazard function ; Translation family ; Power family

Abstract

In 1984 Paul Meier presented an exposition of the Cox regression model at a symposium of the American Society for Artificial Internal Organs. The talk was published the following year (ASAIO Journal, 1985), but received scant attention thereafter. This is a pity: Professor Meier's view of the model remains timely, and is an example of his clear writing style and ability to communicate statistical concepts in a highly informative manner. In this talk I will discuss his view of the Cox model, which began with a description of the survival, density, and hazard rate functions. "In many respects the hazard function is the most intuitive of the three," he wrote, and went on to explain how the hazard function representation is able to accommodate censoring. An interesting aspect of the paper is his contrast of the standard multiple regression model with the Cox model. I will end on this last point with a simulation study illustrating an important issue that was also of concern to Meier: a drawback of the Cox model relative to the standard model is that in a randomized clinical trial adjustment for covariates does not improve the precision of the estimated treatment effect.