Sorry I misspoke, meant to just say independent.
Quote:
|
but maybe it would be OK, or at least good enough for the wiki. Thanks for the help!
|
Well that's exactly the uncertainty the error is supposed to account for. It will be pretty obvious for samples when you just don't have enough data. If you run 2% drop 100 times and get 4 drops, the error bounds you'll generate will be be 4% +/- 3.9% conversely if you ran 1000 trials your error bounds would be down around .7%. It answers the question of if your sample is large enough at the same time.
Useful link
http://en.wikipedia.org/wiki/Binomia...dence_interval What we're doing here is explicitly finding a confidence interval for a binomial test.