PROC LIFETEST - option to request percentiles other than quartiles

[SmithCJGVSU]

PROC LIFETEST will report quartiles of survival times, but it’d be nice to have option to request other percentiles and associated CIs in a option similar in syntax to TIMELIST=.

See:  https://www.linkedin.com/pulse/kaplan-meier-survival-time-percentiles-chris-smith/

[Reeza]

I do agree this would be helpful as an option. 

It's not particularly hard to get at now though, take the output with the survival estimates and generate the percentiles needed. You'll have to make some decisions about rounding, but that's fine. 

I'm not sure how I feel about all percentiles, because they're really estimated by the nearest data point you have not on any distribution, but that's the behaviour right now for the 25/50/75th percentile anyways. 

[SmithCJGVSU]

Thanks for your comment @Reeza.  I agree it’s not particularly difficult programming to post-process the survival estimates table to get the the percentiles, but doing CIs is probably a little more tricky.  Also, it opens up the possibility that a user might derive them incorrectly, I don’t know many people that are aware SAS takes the average of two time points if the survival estimate falls directly on the value of 1 - p.  I work in the pharmaceutical industry where many programmers have backgrounds other than statistics.  I think it would be difficult for a programmer to dig through and understand the details of the computation formulas used in the percentile and CI calculations.  This option would make the process much more user friendly.

[Reeza]

I work in the pharmaceutical industry where many programmers have backgrounds other than statistics.  I think it would be difficult for a programmer to dig through and understand the details of the computation formulas used in the percentile and CI calculations.

Maybe I'm old school, but people working in pharmaceutical should understand survival analysis. Or rely on people who do. These are not particularly complex calculations in a life table or PROC LIFETEST. PROC PHREG is more complicated but LIFETEST is a counting process. 

 I agree it’s not particularly difficult programming to post-process the survival estimates table to get the the percentiles, but doing CIs is probably a little more tricky.

The OUTSURV dataset includes the CI, by default.