Howdy fellow monks,
I am trying to implement the hyper-geometric probability in Perl. The general form of the equation is given, for instance, in the second page of this article, under the heading "Interpretation of Clusters".
Since the values can get quite large, I thought I would implement using Math::Big and Math::BigFloat, but I seem to have a misunderstanding or a bug somewhere that I cannot track down. The code is below. The symptom is that the value of $delta through each loop iteration is 0, and the final probability value is always one. When I do some calculations manually I get distinctly non-zero $delta for the last few loop iterations. Can anyone see what's I've done wrong?
use strict;
use Math::BigFloat;
use Math::Big;
my $G = 40; #$ARGV[0];
my $C = 25; #$ARGV[1];
my $n = 15; #$ARGV[2];
my $k = 14; #$ARGV[3];
sub choose {
my $temp = Math::BigFloat->new('1');
$temp = Math::Big::factorial($_[0]) /
Math::Big::factorial ($_[0] - $_[1]) /
Math::Big::factorial($_[1]);
return $temp;
}
my $p = Math::BigFloat->new('1');
my $denom = Math::BigFloat->new(choose($G, $n));
for (my $i = 0; $i < $k; $i++) {
my $val1 = $G - $C;
my $val2 = $n - $i;
my $delta = Math::BigFloat->new();
$delta = choose($C, $i) * choose($val1, $val2) / $denom;
print "$delta\n";
$p -= $delta;
}
print "Probability estimate: $p\n";
-Tats