Hi all,

Since my last post you'll be happy to know i have successfully built a couple of what appear to be reasonable proportional hazard models

Now i'm trying to asses the fit of these models graphically, i'm used to working with logistic regression models where it is possible to band by predicted propensity and then graph the average predicted propensity and actual propensity in these bands. For Cox regression i've found that this doesn't work... Predicted and actual propensity appear to be very different and i can only imagine this is because of censored data.

I wondered if anyone had any standard graphs they would use to check the fit of these models?

All the best
Laura