topic Re: verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not? in Discussions

verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not?

konradkk — Thu, 24 Jan 2019 18:55:12 GMT

Hi,

I would like to ask if somebody has encounter the issue I got. I would like to verify the 95% CI for a case as below in the table.

95% Confidence Intervals for the True Rate at all Possible Observed Rates within a Subgroup (n=15)

Observed Rate

Lower Bound 95% CI

Upper Bound 95% CI

5/15

0.152

0.723

I used JMP Analyze/Distribution and got the correct Lower CI (0.151763 which verifies 0.152 in the table above) but not Upper (0.582865 vs 0.723) - why would it be?

Level	Count	Prob	Lower CI	Upper CI	1-Alpha
No	10	0.66667	0.417135	0.848237	0.950
Yes	5	0.33333	0.151763	0.582865	0.950
Total	15

Thank you in advance for any help,

Konrad

Re: verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not?

MRB3855 — Mon, 28 Jan 2019 14:22:17 GMT

How are you calculating the numbers in the top table? JMP, in the bottom table, uses the Score method.

Re: verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not?

konradkk — Mon, 28 Jan 2019 17:16:22 GMT

The top numbers were given to me and I only know that SAS (no code was given) was used to derive them and I wanted to verify them using JMP.
Thank you
Konrad

Re: verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not?

MRB3855 — Tue, 29 Jan 2019 14:56:03 GMT

There are a variety of statistical methods, within SAS' proc freq, to calculate such intervals. I *think* the defaults are the Clopper-Pearson and Wald, respectively. But there are many other methods within proc freq (and perhaps some other procs), and without knowing which was used it would be hard to verify via JMP.

https://support.sas.com/documentation/cdl/en/statug/63347/HTML/default/viewer.htm#statug_freq_a0000000660.htm

Re: verifying 95% exact binomial confidence in JMP - Lower CI OK but Upper not?

konradkk — Tue, 29 Jan 2019 15:18:32 GMT

Thank you very much. I requested the SAS code use so hopefully it will help to clarify what method was used. Thank you again, Konrad