topic Re: PROC STDRATE lowerCL missing when direct standardization and gamma CL in Statistical Procedures

PROC STDRATE lowerCL missing when direct standardization and gamma CL

Master_DiSASter — Tue, 21 Jan 2020 06:08:06 GMT

Hello Community,

I have an odd problem, which I cannot get my head around. I'm calculating standardized rates using PROC STDRATE (see code below). whenever I have a lot of observed events the lower confidence limit is not reported (a missing value is reproted). The problem occurs only:

when the number of observed events is greater than 110'000,
METHOD = DIRECT and CL=GAMMA.

Anny suggestions why this problem occurs and how it can be handled?

Thanks a lot in advance and a nice day to everyone!

Kind regards,

Reto

P.S.: Because the problem is connectes to the number of observed events, I cannot provide some sample data.

ods exclude all;
proc stdrate 
data=work.ms_akut
refdata=work.resp 
     method=direct 
     cl=gamma 
     stat=rate(mult=1000)
;
population 
     event=obs_type 
     total=pop 
;
reference 
     total=pop_resp
;
strata sex age_class ;
by canton   
ods output StdRate=output_stdrate;
run;
ods exclude none;

Re: PROC STDRATE lowerCL missing when direct standardization and gamma CL

StatDave — Wed, 22 Jan 2020 14:10:38 GMT

For some data, the gamma-based confidence limits cannot be derived. In these cases a limit might be shown as missing. There should be a message about this in the SAS log. You might consider using a different distribution in the CL= option.

Re: PROC STDRATE lowerCL missing when direct standardization and gamma CL

Master_DiSASter — Wed, 22 Jan 2020 14:51:23 GMT

Thanks a lot for your response.

Could you give more insight why "gamma-based confidence limits cannot be derived" for some data? Is there a mathematical or a convergence problem? Is there a connection to the number of observed events?

With reference to recent literature I would prefer to us the gamma method in connection with directly standardized rates. See for example:

https://pophealthmetrics.biomedcentral.com/articles/10.1186/s12963-018-0177-1

Thanks again and have a nice day,

Reto