Terry Therneau wrote: > > It is easier to get survival curves using the predict function. Here is a > simple example:
>> tfit <- survreg(Surv(time, status) ~ factor(ph.ecog), data=lung) >> tdata <- data.frame(ph.ecog=factor(0:3)) >> qpred <- predict(tfit, newdata= tdata, type='quantile', p=1:99/100) >> matplot(t(qpred), 99:1/100, type='l') > Many thanks - that worked at treat... (One day I might work out what it does - for now I'm happy it does it!) In terms of when I write up what I did is this still a weibull regression? help(predict.survreg) just calls it a quantile... (Sorry that may be dumb question ;-) ) > The above fit assumed a common shape for the 4 groups, > you can add a "+ strata(ph.ecog)" term to have a separate scale for each > group; > this would give the same curves as 4 separate fits to the subgroups. Any thoughts on which is scientifically more valid? I'd have thoughts 4 separate shapes? Certainly if I'm modeling drugs - its surely possible that a new drug might change the course of disease and therefore the shape of the curve altogether? Brings me back to my extra question - is there any way to determine quality of the fit for this (like an R^2 value for a linear regression). That might answer if a strata approach is needed. ______________________________________________ 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.
