topic Re: PROC PHREG (Fine and Gray model) in Statistical Procedures

PROC PHREG (Fine and Gray model)

Alddle — Fri, 18 Nov 2016 03:07:06 GMT

Dear all,

I used proc phreg to run fine and gray model. My dataset has no missing value, and when the univeriate analysis was taken, everything is OK (the number of used observations = the number of read observations). But when the multivariate model was analized, things would be like this, the log as below:

ods graphics on;
proc phreg data=model_os_m plots(overlay=stratum)=cif;
class A (ref='Normal');
class B(order=internal ref=first) C(order=internal ref=first);
model dftime*status(0)=A B C D / eventcode=1;
Hazardratio 'Pairwise' A / diff=pairwise;
baseline covariates=cov_os2_m out=outos2m cif=_all_;
run;

NOTE: 331 observations were deleted due either to missing or invalid values for the time,
censoring, frequency or explanatory variables or to invalid operations in generating the
values for some of the explanatory variables.
NOTE: Convergence criterion (GCONV=1E-8) satisfied.

I have tried proc print/frequency in the Fine and gray model, but the log states that NOTE: The RESMART= option (OUTPUT statement) is ignored for the Fine and Gray competing-risks analysis.

So, why 331 observations have been deleted in the multivariate analysis? And how can I identify which observations have not been used in the final model? Could you please tell me how can I solve this problem in fine and gray model test?

Thank you very much!

Alddle

Re: PROC PHREG (Fine and Gray model)

Reeza — Fri, 18 Nov 2016 05:43:03 GMT

It does say invalid values. Do all your times make sense?

Re: PROC PHREG (Fine and Gray model)

Alddle — Fri, 18 Nov 2016 06:20:13 GMT

Dear Reeza,

Thanks. Do you mean tied event times? Or you mean the robust covariance matrix? But, why univeriate analysis of each factor for fine and fray model is OK? Could you please tell me how can I find these invalid times?

Best,

Alddle

Re: PROC PHREG (Fine and Gray model)

Alddle — Fri, 18 Nov 2016 06:40:38 GMT

I have checked the dftimes, and found that only 61 pairs of observation have tied event times.

Re: PROC PHREG (Fine and Gray model)

Alddle — Wed, 23 Nov 2016 13:14:41 GMT

Thanks. Some observations have been deleted due to a irrelevant variable with missing data uncorrectly involved in the "class" statement.