The likelihood ratio test for strata homogeneity is questionable since some strata have no events.

[Stephen_Q]

When I do survival analysis，all strata have events, but why still got this warning: "The likelihood ratio test for strata homogeneity is questionable since some strata have no events." ?

Log as below: 

Output Added:
-------------
Name: HomTests
Label: Homogeneity Tests
Template: Stat.Lifetest.HomTests
Path: Lifetest.ByGroup3.StrataHomogeneity.HomTests
-------------
WARNING: The likelihood ratio test for strata homogeneity is questionable since some strata have no events.

My code as below, in my dataset cnsr value inclued (0 1), and all trt01an have events(0 1). Why still got this warning? What the meaning?

Does it affect other analyses like ProductLimitEstimates, Quartiles etc. ? Is this survival analysis still valid?

proc lifetest data=adtte timelist=(0 to 20 by 2) atrisk outsurv=surviv reduceout;
by TYPEN;
time aval*cnsr(1);
strata trt01an;
run;

[Ksharp]

Maybe 'aval' has some MISSING value. SAS would delete the entire row if any one of 'aval' 'cnsr' have a missing value .
Using PROC FREQ to check whether each treatment have event or not:

proc freq data=adtte;
where aval is not missing;
table TYPEN*trt01an*cnsr/list;
run;

[Stephen_Q]

Thanks for your reply, 'aval' and 'cnsr' both have no missing value, but some 'TYPEN' dosn't has all   trt01an*cnsr, is that what's causing this? If so how to solve it？

[Ksharp]

"but some 'TYPEN' dosn't has all trt01an*cnsr, is that what's causing this? If so how to solve it？"
Yes. That is where the problem is.
Just keep the validate(integerate) combination of trt01an*cnsr.
a.k.a. at least two levels in trt01an and two levels in cnsr .

[Stephen_Q]

For example，trt01an value inclued(1 2), cnsr value inclued(0 1). I mean some 'TYPEN' only has (trt01an=1 and cnsr=0) ,(trt01an=1 and cnsr=1) ,(trt01an=2and cnsr=0) , dosn't has (trt01an=2and cnsr=1), right?

If so, but in aother dataset, it's a similar situation, The log has no warning.

[Ksharp]

Yes. Delete those records and run the code again. If there is not any warning ,that is to say where the problem is. If not then better post sample data ,so can test your code.