How do i write code and output datasets that include quartiles and pvalues using proc lifetest?

[Hello_there]

Hi, i'm doing a survival table for work and i'm having trouble with the syntax.

I'm trying to use PROC LIFETEST to produce quartiles and p values using a log rank test that is stratified by TRT (treatment) and loc (location). I'm using Kaplan-Meier estimates.

So far this is what i have, but it's not producing what i need. Can i receive some help?

ods output quartiles=a;
proc lifetest data= mydata method=pl;
 time aval*cnsr(1);
 strata trt loc / test=logrank;
run;
ods output close;

[Reeza]

A bit old but possibly useful.

https://gist.github.com/statgeek/d3bce2a9e2ef0523db9d

Something else that may be useful:

https://communities.sas.com/t5/SAS-Communities-Library/MVMODELS-a-Macro-for-Survival-and-Logistic-An...

[Ksharp]

/*
Maybe your dataset have problem.
Can you post your dataset?
I test your code with SASHELP.BMT, 
and have no problem .
*/
proc lifetest data= sashelp.bmt method=pl;
 time t*status(0);
 strata group / test=logrank;
run;

[Hello_there]

Thanks, i think my set up is wrong as i'm trying to output p values.

[Quentin]

Hi, you say:

---

So far this is what i have, but it's not producing what i need. Can i receive some help?

It sounds like your code is running without errors or warnings, is that correct?  Can you say more about how the results you are producing differ from the results you need?  

[Hello_there]

i was trying to produce p values and output them with homtests= dsnout option. There are 4 treatment groups, with the first 3 being active, and the last one being placebo.

The output was only producing one p value, even though i needed 3 since i'm comparing each group to placebo.

[Quentin]

Before worrying about output datasets, I would focus on getting the output results you want.  

When you run the PROC LIFETEST step, do you see the p-values you want?  When I run the step on sashelp.bmt, I get one p-value for a test of equality over strata.  There are 3 groups, but I get one p-value testing overall homogeneity.  Sounds like that's not what you want.  After you figure out how to get the output results you want, it should be possible use ODS to write the results to an output dataset.

[Hello_there]

I think the best option would probably be to do 3 separate tests with the 4 groups, and to use a WHERE statement for the following combos:
where trt in (1,4)
where trt in (2,4)
where trt in (3, 4)
1:3 being active groups,
4 being placebo

That would produce a p value for all the pairwise comparisons against placebo.