[
https://issues.apache.org/jira/browse/MATH-1086?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13868406#comment-13868406
]
Thomas Neidhart commented on MATH-1086:
---------------------------------------
Actually, the BinomialConfidenceInterval class would also be suited for
sub-classing, as all the getXXXInterval methods have the same signature. The
interface would be like this
{noformat}
public interface BinomialConfidenceInterval {
ConfidenceInterval createInterval(int numberOfTrials, int numberOfSuccesses,
double confidenceLevel);
}
{noformat}
And there would be an implementation for each method:
* Wilson
* NormalApproximation
* ...
Maybe an overkill, but at least very flexible ;-)
> Incorrect results from BinomialConfidenceInterval#getWilsonScoreInterval
> ------------------------------------------------------------------------
>
> Key: MATH-1086
> URL: https://issues.apache.org/jira/browse/MATH-1086
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 3.3
> Reporter: Siarhei Khamenka
> Fix For: 3.3
>
>
> It looks like BinomialConfidenceInterval#getWilsonScoreInterval produces not
> accurate results for interval lower bound.
> E.g. for input (10,9,0.95) it returns 0.818 instead of 0.596;
> for input (10,10,0.95) it returns 0.856 instead of 0.722.
> Used also
> http://epitools.ausvet.com.au/content.php?page=CIProportion&SampleSize=10&Positive=9&Conf=0.95&Digits=3
> and http://www.measuringusability.com/wald.htm to verify it.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
