Cox Predicted Survival
This is a slightly modified version of Therneau's
function. The difference is that
survfit.cph assumes that
x=TRUE,y=TRUE were specified to the fit. This assures that the
environment in effect at the time of the fit (e.g., automatic knot
estimation for spline functions) is the same one used for basing predictions.
## S3 method for class 'cph': survfit(formula, newdata, se.fit=TRUE, conf.int=0.95, individual=FALSE, type, vartype, conf.type=c('log', "log-log", "plain", "none"), ...)