Re: [boost] Fisher Exact test in boost ?

Subject: Re: [boost] Fisher Exact test in boost ?
From: John Maddock (boost.regex_at_[hidden])
Date: 2010-07-23 12:15:04

> I will try to perform the implementation myself using the updated 
> hypergeometric.
>
> Has anyone pointers to efficient algorithm for two-sided 2x2 Fisher  exact 
> (or Barnard) test, able to deal with mix of small  and large  cell counts 
> (  0 and ~ 10,000) ?
> Large cell counts are typically a problem for the classical factorial 
> version.

Our hypergeometric implementation will cope with the large factorials OK, 
but you're correct that it may be slow for N > 170.

John.