I have already done a survival analysis using cox regression on the churn dataset to find out the optimal parameters for getting the maximal survival time in my dataset. I have got a decent curve which represents the Mean Time of Survival for my dataset.But I am told to come up with a plan or code in R such that I can predict the estimated survival time on a test dataset that would predict my survival time. Is it possible to use the Cox regression model to do this?
Asked
Active
Viewed 419 times
0
-
Is this a self-study question? Look at `R`'s `predict()` family of functions. – Todd D Sep 25 '18 at 15:04
-
A useful answer [here](https://stats.stackexchange.com/questions/55118/cox-model-vs-logistic-regression/55434#55434). Basically: the Cox model doesn't directly estimate a survival curve. You have to do some advanced statistical procedures. [The Breslow estimator](https://www.stat.washington.edu/jaw/RESEARCH/TALKS/NBws.pdf) obtains a stepwise approximation to the baseline cumulative hazard function which can be exponentiated using the Cox coefficients to obtain the predicted survival. – AdamO Sep 25 '18 at 15:06