Partial Hazard interpretation in CoxTimeVaryingFitter

[EleniNeti]

Hi everyone,
I have brought my dataset in the long format as suggested by the documentation. Problem arises when I do the predict_partial_hazard method, since I find it hard to interpret the output precisely. Does each prediction refer to each patient? I'm confused since there can be multiple instances referring to the same patient ID. Can someone that used it before give some help regarding the interpretation?

[AhmetZamanis]

Hi,

The predict_partial_hazard method will give you the estimated partial hazard of each observation, based on their covariate values. If a patient has multiple observations / measurements in your dataset, with different covariate values, the estimated partial hazard be different for each observation.

How you interpret that depends on what you need to analyze. In a project, I plotted the estimated partial hazard values against covariate values to try & understand how the covariates affect the partial hazard estimates.