Welcome to the hotseat. We've prepared a guide if you'd like to read more about how it works.

Exploratory data analysis for survival analysis (Cox models)

0 votes
asked Apr 6, 2016 in BUS 3018F - Models by carapienaar (200 points)
Is there anything you should be looking out for in your data before fitting a Cox hazard model? 

For example, with other kinds of models, like linear regression, you should check if your response is normally distributed. I'm wondering if it's the same for Cox proportional hazard models - if any variables need to follow a particular distribution or if there are any other important assumptions. 

If so, what are common ways of finding and dealing with problems in this kind of data? 


1 Answer

+2 votes
answered Apr 11, 2016 by JasonWessels (330 points)
selected Apr 18, 2016 by carapienaar
Best answer

The Cox regression model is a semi-parametric model in the sense that you do not need to specify a functional form of the baseline hazard rate. So you do not need to assume that any variables follow a particular distribution therefore your response does not need to follow any particular distribution (you can however fit a fully parametric proportional hazards model by assuming that the survival times follow a specific distribution and then checking to see how well that distribution fits the data). The main thing you need to look out for in the data is any truncation or censoring of observations.