Exact Binomial CI for Proportion Difference - PROC FREQ

[Doug____]

What methods are available if there are no responses in a 2x2 table. For example, when analyzing proportion differences between treatments by Gender, neither Male nor Female have any responses (e.g. the counts are zero for both), so FREQ will not estimate the CIs for the difference in proportion. To get around this, I placed a very small non-zero value in the counts table, say 0.000000001 for the zero response counts and then processed it with FREQ using EXACT and RISKDIFF. Should some other method be used, and if so, what is it?

[SAS_Rob]

Unfortunately, PROC FREQ doesn’t provide RISKDIFF analysis for two-way tables when there is a zero-frequency column(both treatment arms have no positive responses). In the meantime the work-around to assign very small weights to the zero-frequency cells in your two-way table is the only option. When you use this work-around, the following confidence limit types are available for the risk difference: Agresti-Caffo, Hauck-Anderson, Miettinen-Nurminen (score), and Newcombe. Of course the traditional/Wald confidence limits are also available, but equal (0,0) in the case of a zero-frequency column where the estimate of the risk difference is 0.   

[Doug____]

Alright so are the CLs obtained through an ODS table or through an output statement? I tried an output statement but the columns are not easily discerned.

In addition, the small weights do not appear to work as the CI for the risk difference column where the zero weights occur are still showing missing values. Below shows the code I used after applying the small weights.

proc freq data = countsfreq;
by subgroupan subgroupa group ;
table trtp*ev /out = x riskdiff (cl=newcombe) nowarn alpha = 0.1;
weight count/zeros;
/* exact riskdiff;*/
/* ods output riskdiffcol2 = ci1 ;*/
output out = rd2 riskdiff2;
run;

If I use the riskdiff statement my version of SAS only allows (method = score) and with the options above it produces no CI values for the items with zero responses.