Plot bootstrapped survival curves

[DF Moreno]

Dear Statalist,

Is there a way to plot multiple survival curves after using a Cox model with bootstrap? I mean, to plot 100 (or any other) replicates from bootstrap in a single Kaplan-Meier graph.
I was thinking of the next pseudocode:

* stset date2, failure(event==1) origin(date1) id(id)
* stcox my_variable, vce(bootstrap, reps(100) seed(1234))
* sts gen s_myvar = f, by(my_variable)
* sort _t
* twoway connected s_myvar _t

I do not know how to integrate the last three lines to get the estimates from each of the 100 bootstrap replicate so I can plot them into a single Kaplan-Meier.
Thank you!

[Tiago Pereira]

Hi, DF.

Your problem was explained vaguely. I am not sure if this is the best solution (you must check if it makes sense from a statistical point of view. I am not familiar with bootstrapping time-to-event outcomes), but (initially) I would approach your problem as follows:

1. Generate the main survival curve and save it.
2. Loop N times (for i = 1,...,100), every time getting a bootstrapped sample.
3. For each iteration (boostrap), fit the Cox model
4. Save the survival function in a temporary dataset:

 stcurve, survival at1(covariate=1) outfile(iteration`i', replace)

5. Append the N saved survival curves, identifiying them with an ID
6. Plot the results via -spagplot-
7. Highlight the main survival curve (which I believe should be plot 1)

Hope this gives you some direction.

Of course you will have to adapt the algorithm above, since you have two or more groups/exposures.