0
$\begingroup$

Lets suppose I have aggregate survival probabilities (NOT individual level data) from Kaplan Meier curves. I would like to curve fit different "types" of distribution like exponential, weibull, log-normal, and log-logistic on this curve so that I can extrapolate.

Below is the sample data derived from KM for aggregated data.

surv <- c(0.92,0.86,0.80,0.77,0.73,0.70,0.68,0.63,0.61,0.59,0.57)

enter image description here

Below are my questions:

  1. How would I curve fit the data to the aforementioned distributions, to derive parameters such as scale and shape in case of weibull?
  2. Any illustrative examples in Excel or R would very helpful.
$\endgroup$
  • $\begingroup$ Is there any reason that you can't use a generic non-linear curve fitting function, like nls in R? $\endgroup$ – EdM May 30 '15 at 16:16
  • $\begingroup$ @EdM, non-linear curving fitting algorithm always provide illogical curves, a better fit would be to fit an assumed probability distribution based on domain knowledge. $\endgroup$ – forecaster May 31 '15 at 2:01
  • $\begingroup$ The survreg function in R's survival package will do this. (looks like you are already familiar with the syntax at least on some level.) $\endgroup$ – DWin Jun 24 '15 at 4:17

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.