Thanks for your reply.
I'm trying to use "predictProb.coxph()" function in "peperr" package to
calculate the predicted probability.
Here the particular time, do you mean the total follow-up time for each
subject? Because the predict.coxph() function didn't provide a time parameter
to input, I can't get the expected events number at a given time point,i.e.
2-year. And for the cumulative hazard, how is it calculated for each subject?
Because the final Cox model only gives the estimated coefficients, does it
only calculate the linear predictor or risk score? Another question is whether
this "expected" vector can be used to evaluate the Cox model, such as C-index,
ROC ?
Sorry for so many questions, I don't understand Cox very well,Thanks
> Date: Fri, 30 Mar 2012 08:48:25 -0500
> From: thern...@mayo.edu
> To: jgz02...@hotmail.com
> Subject: Re: how to calculate predicted probability of Cox model
>
> -- begin included message ---
> Because Cox proportional hazards model didn't give the baseline hazard
> function, how to calculate the predictive probability for each test
> sample at a special time point,such as 5-year or 10-year ?
> In survival package, predict.coxph() function gives three different
> type of predicted value, for the type of "expected", Does it mean the
> expected number of events for a special sample during the total
> follow-up time? What's the relationship between this value and
> predicted probability?
> -- end inclusion ---
>
> The default printout for coxph does not include the baseline hazard (too
> lengthy), but it is lurking behind the scenes.
>
> Yes, the type "expected" is the expected number of events (= value of
> the cumulative hazard) for a given set of covariates at a particular
> time. Prob(event by time t) = exp(- cumulative hazard at time t).
> To get a prediction of type = expected you data set has to have both
> covariate AND survival time values for each subject.
>
> Terry T.
>
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
