Solved – the graph of log(-log) for Cox model on survival analysis

I'm studying Cox Regression model on Survival Analysis.

While testing validity of Proportional Hazard model, I will use log(-log) graph method in SPSS.

First of all, I mention which procedure I'm following using SPSS.

I select Analyse –> Survival –> Kaplan-Meier –> defined time, status, factor. And then I select Save option and tick "Survival". "Group" variable is assigned as factor. And I get the following output "Sur_1". How can I fulfil the blank cell in the variable "Sur_1" ? If I'm able to fulfil them, I will compute log(-log Sur_1) and I plot it. Please help me thank you.

enter image description here
enter image description here

The observations with the blanks would appear to be the censored observations. With KM/Cox type approaches the estimated survival function does not change at these times so you can ignore them in the plotting.

A log (-log S) plot does not per-se tell you the validity of the model, it can only help to illustrate whether the hazards are approx. proportinal throughout. Even if that is not the case, you still obtain a reasonablle test (more or less the logrank test depending on how you handle tied event times) and some people would argue that the estimated hazard ratio still has an "average" interpretation – even if that is certainly more complicated, of more debatable relevance and harder to interpret.

Similar Posts:

Rate this post

Leave a Comment