Lectures and seminars Faculty event: Paul Dickman

Title

On the proportional hazards assumption in Cox regression

Abstract 

The focus of this presentation is discussion of the 2020 paper by Mats Stensrud and Miguel Hernán "Why Test for Proportional Hazards?", in particular their statement that ‘statistical tests for proportional hazards are unnecessary’. I don't disagree with this statement, or anything else in their paper, but think there are interesting points to discuss. For example, I am concerned that the statement may be (mis)interpreted by some as ‘assessing the proportional hazards assumption (and implications for inference) is unnecessary’. I will start by giving a brief overview and definition of the proportional hazards assumption. 'Proportional hazards' can be viewed as 'an assumption of no interactions between covariates and time'. Before discussing the statement that 'statistical tests for this specific interaction are not required' I will discuss the broader view of null hypothesis significance testing in epidemiology and also the role of assumptions, in general, in statistical modelling in epidemiology. Finally, I will explore whether the statement that ‘statistical tests for proportional hazards are unnecessary’ can be extended to 'statistical tests for interactions are unnecessary'.

References

Why Test for Proportional Hazards?
Stensrud MJ, Hernán MA
JAMA 2020 Apr;323(14):1401-1402

Why test for proportional hazards - or any other model assumptions?
Sjölander A, Dickman P
Am J Epidemiol 2024 Feb;():

About the speaker 

Paul Dickman is Professor of Biostatistics at MEB, where he has been employed since March 1999. His primary research interests are developing and applying statistical methods for population-based studies of cancer patient survival, particularly the estimation and modeling of relative/net survival. He has general interests in register-based epidemiology, especially cancer epidemiology. He is head of the biostatistics group at MEB and programme director for the master's programme in biostatistics and data science.

Join the faculty event in Wargentin! After the seminar, the faculty lunch will be served in Ljusgården.