Difference of two proportions with 95% CI

[mariko5797]

I need to find the difference of proportions between two treatment groups and obtain their 95% CI via Clopper-Pearson methodology.

Is there a way to use a SAS procedure, such as PROC FREQ, to obtain the 95% CI or do I need to manually calculate?

The following code is what I've tried to use to obtain CI, but none are being outputted.

data have;
 label trt= 'Treatment' 
 	   param= 'Parameter'
 	   response= 'Response 1=success 2=fail'
	   x= 'Counts'
	   n= 'Total'
	   ;
 input trt param response x n @@;
 cards;
 	1	1	1	9	13
	1	2	1	13	13
	1	3	1	6	13
	1	1	2	4	13
	1	2	2	0	13
	1	3	2	7	13
 	2	1	1	17	18
	2	2	1	12	18
	2	3	1	14	18
	2	1	2	1	18
	2	2	2	6	18
	2	3	2	4	18
	;
proc sort; by param;
proc freq noprint;
 by param;
 tables trt*response / riskdiff binomial(CL= clopperpearson) out= dsnout1;
 weight x;
 output binomial n out= dsnout2;
run;

[ballardw]

When you use NOPRINT then you told the procedure not print the output for anything. The OUTPUT statement would need to include options related to relative risks if Rel1, Rel2, Riskdiff, Riskdiff1, Riskdiff2, etc, Look at the online help under output, there's about a dozen related.

Or don't use NOprint and do use the ODS output to capture the entire Riskdiff output:

ods output RiskDiffCol1= RisksColumn1
           RiskDiffCol2= riskscolumn2
;

proc freq data=have;
 by param;
 tables trt*response / riskdiff binomial(CL= clopperpearson) out= dsnout1;
 weight x;
 output binomial n out= dsnout2;
run;

[Ksharp · Accepted solution]

What kind of proportions are you refering to ?

If you mean x/n , then check this:

http://support.sas.com/kb/24/188.html

If you mean Response=1 's ratio(the number of Response=1/ the number of each treat group), check PROC FREQ's relrisk:

proc sort; by param;
proc freq ;
 by param;
 tables trt*response / riskdiff(column=1 CL=exact) ;
 weight x;
 output out= dsnout2 riskdiff;
run;