topic Re: Help to understand reliability analysis in Statistical Procedures

Help to understand reliability analysis

mikemangini — Wed, 21 Oct 2015 13:28:45 GMT

SAS Community, I found this code in an archive at work and it contains no comments. The idea is to calculate a reliable sample size. It is not working properly and I am seeking to understand each component.

The input table cy11rates contains the unique provider ID, risk adjusted rate of complication (rsrr_hwr) and total patient volumn. Is anyone familiar with this process that could help me understand?

data temp;
set cy11rates;
numerator=round(rsrr_hwr*volume);
do i=1 to volume;
d=1;
if i<= numerator then n=1;
else n=0;
output;
end;
run;

proc sql;
select sum(volume) into:cnt
from cy11rates;
quit;

proc freq data=temp;
tables provid /out=b;
run;

data c;
set b;
rtsq=count*count;
rtdv=rtsq/&cnt.;
run;

proc summary sum data=c;
var rtdv;
output out=new sum=nv;
run;

data new2;
set new;
t=(&cnt.-nv)/(_freq_-1);
run;

proc print data=new2; run;

proc anova data=temp;
ods output OverallANOVA =anova_result;
class provid;
model n=provid;
run;
quit;

proc sql;
create table reliability as
select distinct (Fvalue-1)/(Fvalue-1+t) as ICC,
t*(calculated ICC)/(1+(t-1)*(calculated ICC)) as reliability,
(0.7-0.7*(calculated ICC))/((calculated ICC)-(0.7*(calculated ICC))) as sample_size
from new2
join anova_result
on 1=1
where anova_result.source='Model';
quit;

Re: Help to understand reliability analysis

ballardw — Wed, 21 Oct 2015 14:38:36 GMT

Please describe how "it is not working properly". Error messages? Post the errors and log.

No output? Unexpected output? How is it unexpected.

Also dealing with sample size generation I would recommend looking into proc power.

Re: Help to understand reliability analysis

mikemangini — Wed, 21 Oct 2015 14:41:49 GMT

Thank you for your reply. I received a sample size that was a negative number. My hope was to better understand the workings of the procedure so I could determine if adjustments are needed or if the code is bad. The hope was to get a general idea of each step. I will look into proc power though and appreciate that suggestion.

Re: Help to understand reliability analysis

mikemangini — Wed, 21 Oct 2015 14:48:11 GMT

My understanding of proc power is to plan sample size for a comparison of means. In my analysis I do not have an intervention. I wish to know the minimum patient volume for a provider group so that a risk adjusted complication rate can be considered stable.