03-27-2016 12:49 PM - last edited on 03-27-2016 02:02 PM by Reeza
DATA covprob; p = .5; q = 1-p; z=abs(probit(.025)); zz=z*z; do n = 10 to 500; w = zz / (2 * n); rad = sqrt(2*w*p*q + w*w); rt = (p + w + rad)/(1 +2*w); lt = (p + w - rad)/(1 +2*w); nhat_rt = floor(n*rt); nhat_lt = ceil(n*lt - 1); ncurl_rt = floor((n+4)*rt - 2); ncurl_lt = ceil((n+4)*lt - 3); p1 = probbnml(p, n, nhat_lt); p2 = probbnml(p, n, nhat_rt); cv_hat = p2 - p1; p3 = probbnml(p, n, ncurl_lt); p4 = probbnml(p, n, ncurl_rt); cv_curl = p4 - p3; output; end; keep n cv_hat cv_curl; run; PROC plot data = covprob; plot cv_hat*n = '^'; plot cv_curl*n = '~'; run; quit;
03-27-2016 02:04 PM
In what context? Is there a specific part you don't understand?
Are you looking for an explanation of a specific line?
Where did you find this code and what are you trying to do with it?
03-27-2016 03:31 PM
It looks like some sort of power analysis for (binomial) proportions. But why reverse engineer this undocumented code? Understand the problem and write your own. And document it!