blib/lib/Statistics/MVA/Bartlett.pm

filter: Statementstmt 26.7% 23 / 86 Branchbran 0.0% 0 / 6 Conditioncond n/a MC/DCmcdc n/a Subroutinesub 53.8% 7 / 13 Podpod 0.0% 0 / 1 totaltotal 28.3% 30 / 106 CC 10 SCAR 19.3
CC
10
Cov
25%
_final
12.0
_cv_matrices
12.0
_p_notall
12.0
CRAP 6.9
↑ Summary
1 package Statistics::MVA::Bartlett;
2 1 ▶use base Statistics::MVA;
Subroutine BEGIN: called CC=1 SCAR=0.0
3  
4 1 ▶use strict;
Subroutine BEGIN: called CC=1 SCAR=0.0
5 1 ▶use warnings;
Subroutine BEGIN: called CC=1 SCAR=0.0
6 1 ▶use Carp;
Subroutine BEGIN: called CC=1 SCAR=0.0
7 1 ▶use Math::Cephes qw(:explog);
Subroutine BEGIN: called CC=1 SCAR=0.0
8 1 ▶use Statistics::Distributions qw( chisqrdistr chisqrprob );
Subroutine BEGIN: called CC=1 SCAR=0.0
9  
10 =head1 NAME
11  
12 Statistics::MVA::Bartlett - Multivariate Test of Equality of Population Covariance Matrices.
13  
14 =cut
15  
16 =head1 VERSION
17  
18 This document describes Statistics::MVA::Bartlett version 0.0.4
19  
20 =cut
21  
22 =head1 SYNOPSIS
23  
24 # we have several groups of data each with 3 variables
25 my $data_X = [
26 [qw/ 191 131 53/],
27 [qw/ 200 137 52/],
28 [qw/ 173 127 50/],
29 [qw/ 160 118 47/],
30 [qw/ 188 134 54/],
31 [qw/ 186 129 51/],
32 [qw/ 163 115 47/],
33 ];
34  
35 my $data_Y = [
36 [qw/ 211 122 49/],
37 [qw/ 201 144 47/],
38 [qw/ 242 131 54/],
39 [qw/ 184 108 43/],
40 [qw/ 223 127 51/],
41 [qw/ 208 125 50/],
42 [qw/ 199 124 46/],
43 ];
44
45 my $data_Z = [
46 [qw/ 185 134 50/],
47 [qw/ 171 128 49/],
48 [qw/ 174 131 52/],
49 [qw/ 186 107 49/],
50 [qw/ 211 118 51/],
51 [qw/ 217 122 49/],
52 ];
53  
54 use Statistics::MVA::Bartlett;
55
56 # Create a Statistics::MVA::Bartlett object and pass it the data as a series of Lists-of-Lists within an anonymous array.
57 my $bart1 = Statistics::MVA::Bartlett->new([$data_X, $data_Y, $data_Z]);
58  
59 # Access the output using the bartlett_mva method. In void context it prints a report to STDOUT.
60 $bart->bartlett_mva;
61  
62 # In LIST-context it returns the relevant parameters.
63 my ($chi, $df, $p) = $bart->bartlett_mva;
64  
65 =cut
66  
67 =head1 DESCRIPTION
68  
69 Bartlett's test is used to test if k samples have equal variances. This multivariate form
70 tests for homogeneity of the variance-covariance matrices across samples. Some statistical tests assume such homogeneity
71 across groups or samples. This test allows you to check that assumption. See
72 http://www.itl.nist.gov/div898/handbook/eda/section3/eda357.htm.
73  
74 =cut
75  
76 1 ▶use version; our $VERSION = qv('0.0.4');
Subroutine BEGIN: called CC=1 SCAR=0.0
77  
78 0 ▶sub bartlett_mva {
Subroutine bartlett_mva: not called CC=2 SCAR=17.9
Pod undocumented
79 0 my $self = shift;
80 0 my $context = wantarray;
81 0 my $s_ref = $self->[0];
82 0 my $k = $self->[1];
83 0 my $p = $self->[2];
84
85 0 my $c_ref = &_Cs($s_ref, $k, $p);
86 0 my $d_ref = &_dets($c_ref, $k);
87 0 my $chi = &_final($d_ref, $k, $p);
88 0 my $df = $p*($p+1)*($k-1)/2;
89 0 my $chisprob = &chisqrprob($df,$chi);
90  
91 0 ▶ if ( !$context ) { print qq{\nChi = $chi, df = $df and p = $chisprob}; return; }
Branch: 1 branch
BranchTrueFalseTotal
if (not $context) { } 0 0 0
92 0 else { return ($chi, $df, $chisprob); }
93 }
94  
95 0 ▶sub _dets {
Subroutine _dets: not called CC=2 SCAR=17.9
96 0 my ($c_ref, $k) = @_;
97 # d<-det(c) # d_a<-det(c_a) # d_b<-det(c_b)
98 0 my $det_ref = [];
99 0 for my $i (0..$k) {
100 0 my $determinant = $c_ref->[$i][0]->det;
101 0 $det_ref->[$i] = [$determinant, $c_ref->[$i][1]];
102 }
103 0 return $det_ref;
104 }
105  
106 0 ▶sub _final {
Subroutine _final: not called CC=3 SCAR=24.8
107 0 my ($d_ref, $k, $p) = @_;
108 0 my $d = ( ( $d_ref->[$k][1] - $k ) * ( log ( $d_ref->[$k][0] ) ) );
109 0 my $d_sum = 0;
110 0 for my $i (0..$k-1) { $d_sum += ( ( $d_ref->[$i][1] - 1 ) * ( log ( $d_ref->[$i][0] ) ) ); }
111 0 my $m = $d - $d_sum;
112 0 my $n_sum = 0;
113 0 for my $i (0..$k-1) { $n_sum += 1 / ( $d_ref->[$i][1] - 1 ); } # print 1/($d_ref->[$i][1]-1) }
114 0 my $h = 1 - ( ( 2 * $p* $p+3 * $p-1 ) / (6 * ($p+1) * ($k-1) ) * ( $n_sum - (1/($d_ref->[$k][1]-$k))) );
115 0 my $chi = $m * $h;
116 0 return $chi;
117 }
118  
119 0 ▶sub _Cs {
Subroutine _Cs: not called CC=2 SCAR=17.9
120 0 my ( $s_ref, $k, $p ) = @_;
121 0 my $c_ref = [];
122 #/ matrices are initialised as zero! so we can just use addition... and not shadow or initialise a p*p structure of 0s
123 0 my $c_total = Math::MatrixReal->new($p, $p);
124 # $new_matrix = $some_matrix->shadow();
125 0 my $n_total = 0;
126 0 for my $i (0..$k-1) {
127 0 my $c_mat = Math::MatrixReal->new($p, $p);
128 0 my $factor = (1/($s_ref->[$i][1]-1));
129 0 $c_mat->multiply_scalar($s_ref->[$i][0],$factor);
130 0 $c_total->add($c_total,$s_ref->[$i][0]);
131 0 $n_total += $s_ref->[$i][1];
132 0 $c_ref->[$i] = [$c_mat, $s_ref->[$i][1]];
133 }
134 0 my $factor = (1/($n_total-$k));
135 0 $c_total->multiply_scalar($c_total,$factor);
136 0 $c_ref->[$k] = [$c_total,$n_total];
137 0 return $c_ref;
138 }
139  
140 0 ▶sub _cv_matrices {
Subroutine _cv_matrices: not called CC=3 SCAR=24.8
141 0 my ( $groups, $k ) = @_;
142 0 my @p;
143 0 my $a_ref = [];
144 #for my $i (0..$self->[1]-1) {
145 0 for my $i (0..$k-1) {
146 #my $mva = Statistics::MVA->new($groups->[$i],{standardise => 0, divisor => 1});
147 0 my $mva = Statistics::MVA->new($groups->[$i]);
148 0 my $mva_matrix = my $a = Math::MatrixReal->new_from_rows($mva->[0]);
149 0 push @p, $mva->[1];
150 0 $a_ref->[$i] = [ $mva_matrix, $mva->[2] ];
151 }
152 #/ should make sure than p for all entries is the same!!!
153 0 my $p_check = shift @p;
154 0 ▶ croak qq{\nAll groups must have the same variable number} if &_p_notall($p_check, \@p);
Branch: 1 branch
BranchTrueFalseTotal
if &_p_notall($p_check, \@p) 0 0 0
155 #croak qq{\nAll groups must have the same variable number} if &_p_notall(@p);
156 0 return ($a_ref, $p_check);
157 }
158  
159 0 ▶sub _p_notall {
Subroutine _p_notall: not called CC=3 SCAR=24.8
160 0 my ($p_check, $p_ref) = @_;
161 #for (@p) {
162 0 for (@{$p_ref}) {
163 # return 0 if $_ != $p_check;
164 0 ▶ return 1 if $_ != $p_check;
Branch: 1 branch
BranchTrueFalseTotal
if $_ != $p_check 0 0 0
165 }
166 0 return 0;
167 }
168  
169 1 1; # Magic true value required at end of module
170  
171 __END__
↑ Summary