File Coverage

blib/lib/Text/NSP/Measures/2D/MI/ll.pm

Criterion	Covered	Total	%
statement	24	25	96.0
branch	2	2	100.0
condition			n/a
subroutine	6	7	85.7
pod			n/a
total	32	34	94.1

=head1 NAME

Text::NSP::Measures::2D::MI::ll - Perl module that implements Loglikelihood
                                  measure of association for bigrams.

=head1 SYNOPSIS

=head3 Basic Usage

  use Text::NSP::Measures::2D::MI::ll;

  my $npp = 60; my $n1p = 20; my $np1 = 20;  my $n11 = 10;

  $ll_value = calculateStatistic( n11=>$n11,
                                      n1p=>$n1p,
                                      np1=>$np1,
                                      npp=>$npp);

  if( ($errorCode = getErrorCode()))
  {
    print STDERR $errorCode." - ".getErrorMessage();
  }
  else
  {
    print getStatisticName."value for bigram is ".$ll_value;
  }

=head1 DESCRIPTION

The log-likelihood ratio measures the deviation between the observed data
and what would be expected if  and  were independent. The
higher the score, the less evidence there is in favor of concluding that
the words are independent.

Assume that the frequency count data associated with a bigram
 as shown by a 2x2 contingency table:

          word2   ~word2
  word1    n11      n12 | n1p
 ~word1    n21      n22 | n2p
           --------------
           np1      np2   npp

where n11 is the number of times  occur together, and
n12 is the number of times  occurs with some word other than
word2, and n1p is the number of times in total that word1 occurs as
the first word in a bigram.

The expected values for the internal cells are calculated by taking the
product of their associated marginals and dividing by the sample size,
for example:

          np1 * n1p
   m11=   ---------
            npp

Then the deviation between observed and expected values for each internal
cell is computed to arrive at the log-likelihood value.

 Log-Likelihood = 2 * [n11 * log(n11/m11) + n12 * log(n12/m12) +
           n21 * log(n21/m21) + n22 * log(n22/m22)]

=head2 Methods

=over

=cut


package Text::NSP::Measures::2D::MI::ll;


use Text::NSP::Measures::2D::MI;
use strict;
use Carp;
use warnings;
no warnings 'redefine';
require Exporter;

our ($VERSION, @EXPORT, @ISA);

@ISA  = qw(Exporter);

@EXPORT = qw(initializeStatistic calculateStatistic
             getErrorCode getErrorMessage getStatisticName);

$VERSION = '0.97';

=item calculateStatistic() - This method calculates the ll value

INPUT PARAMS  : $count_values       .. Reference of an hash containing
                                       the count values computed by the
                                       count.pl program.

RETURN VALUES : $loglikelihood      .. Loglikelihood value for this bigram.

=cut

sub calculateStatistic
{
  my %values = @_;

  # computes and sets the observed and expected values from
  # the frequency combination values. returns 0 if there is an
  # error in the computation or the values are inconsistent.
  if( !Text::NSP::Measures::2D::MI::getValues(\%values) )
  {
    return;
  }

  #  Now for the actual calculation of Loglikelihood!
  my $logLikelihood = 0;

  # dont want ($nxy / $mxy) to be 0 or less! flag error if so!
  $logLikelihood += $n11 * Text::NSP::Measures::2D::MI::computePMI( $n11, $m11 );
  $logLikelihood += $n12 * Text::NSP::Measures::2D::MI::computePMI( $n12, $m12 );
  $logLikelihood += $n21 * Text::NSP::Measures::2D::MI::computePMI( $n21, $m21 );
  $logLikelihood += $n22 * Text::NSP::Measures::2D::MI::computePMI( $n22, $m22 );

  return ( 2 * $logLikelihood );
}


=item getStatisticName() - Returns the name of this statistic

INPUT PARAMS  : none

RETURN VALUES : $name      .. Name of the measure.

=cut

sub getStatisticName
{
    return "Log-likelihood";
}



1;
__END__