I'm doing survival analysis with a data set of 200 patients. They each have 5 continuous biomarker variables, a time variable and a yes/no event variable. I fit a Cox proportional hazards model with the 5 biomarker variables and I'm using the model's linear predictor as a prognostic index (PI); each patient has a PI value and values range from -10 to +2.
I'd like to convert the PI values to absolute risks (eg what is the probability of the event if the PI=-7, if the PI=-1 etc) so that I can plot absolute risk by PI and see how absolute risk changes as the PI changes.
I know that a Cox model is a relative risk model where the baseline risk is unspecified, so I'm stuck on how to: 1) Get a value for this baseline hazard , and 2) How to combine this baseline hazard with my PI (ie the Cox linear predictor) to get absolute risks
