Calculating CI %

[Vijay77]

Is it possible to calculate Clopper-Pearson CIs for Proportions of zero.

I have 0 obs in output, but i have created a dummy like with count of 0.

var     trt        count

A        trt1       0

B        trt2       0

C        trt3       0

Is it possible to calculate Clopper-Pearson CIs for Proportions of zero. I only require upper and lower limits per trt.

Thanks,

vijay

[ed_sas_member]

Hi @Vijay77 

Here seems to be a useful link: https://www.phusewiki.org/docs/2015_Lonavala_SDE/Presentations/Clopper-Pearson%20Confidence%20Interv...

[FreelanceReinh]

Hello @Vijay77,

---

@Vijay77 wrote:

var     trt        count

A        trt1       0

B        trt2       0

C        trt3       0

Is it possible to calculate Clopper-Pearson CIs for Proportions of zero. I only require upper and lower limits per trt.

---

Yes, this is possible, but you need the group totals for the calculation. I have mocked up the missing data in the example below (25 subjects in treatment group 1, 30 in group 2 and 40 in group 3).

data have;
input var $ trt $ count;
cards;
A trt1 0
X trt1 25
B trt2 0
Y trt2 30
C trt3 0
Z trt3 40
;

proc freq data=have;
by trt;
weight count / zeros;
tables var / binomial(level=1);
run;

Without the ZEROS option of the WEIGHT statement PROC FREQ would ignore the observations with COUNT=0. You find the 95%-Clopper-Pearson CIs in section "Exact Conf Limits" of the output (for each BY group), as usual. I leave it to you to make adjustments for multiplicity, if desired.