blib/lib/Statistics/RankCorrelation.pm

filter: Statementstmt 98.6% 147 / 149 Branchbran 87.5% 49 / 56 Conditioncond 70.8% 17 / 24 MC/DCmcdc 42.8% 6 / 14 Subroutinesub 100% 19 / 19 Podpod 100% 16 / 16 totaltotal 91.3% 254 / 278 CC 51 SCAR 17.9
CC
51
Cov
93%
_init
12.3
kendall
10.0
spearman
6.0
CRAP 6.0
↑ Summary
1 package Statistics::RankCorrelation;
2 2 our $AUTHORITY = 'cpan:GENE';
3 # ABSTRACT: Compute the rank correlation between two vectors
4  
5 2 ▶use strict;
Subroutine BEGIN: called CC=1 SCAR=0.0
6 2 ▶use warnings;
Subroutine BEGIN: called CC=1 SCAR=0.0
7  
8 2 our $VERSION = '0.1206';
9  
10 15 ▶sub new {
Subroutine new: called CC=2 SCAR=7.1
Pod documented
11 15 my $proto = shift;
12 15 ▶ my $class = ref($proto) || $proto;
Condition33%1/3
ref $proto || $proto
l!l&&r!l&&!r
0150
MC/DC0%0/2
ref $proto || $proto
ref $proto $proto
Truth table33%1/3
A || B
A= ref $protoB= $proto
ABresult
1X1
011
000
13 15 my $self = {};
14 15 bless $self, $class;
15 15 $self->_init(@_);
16 15 return $self;
17 }
18  
19 15 ▶sub _init {
Subroutine _init: called CC=12 SCAR=25.1
20 15 my $self = shift;
21  
22 # Handle vector and named parameters.
23 15 ▶ while( my $arg = shift ) {
Branch: 1 branch
BranchTrueFalseTotal
if (my $arg = shift()) 29 15 44
24 29 ▶ if( ref $arg eq 'ARRAY' ) {
Branch: 2 branches
BranchTrueFalseTotal
if (ref $arg eq 'ARRAY') { } 28 1 29
elsif (not ref $arg) { } 1 0 1
25 28 ▶ if( !defined $self->x_data ) { $self->x_data( $arg ) }
Branch: 2 branches
BranchTrueFalseTotal
if (not defined $self->x_data) { } 14 14 28
elsif (not defined $self->y_data) { } 14 0 14
26 14 elsif( !defined $self->y_data ) { $self->y_data( $arg ) }
27 }
28 elsif( !ref $arg ) {
29 1 my $v = shift;
30 1 ▶ $self->{$arg} = defined $v ? $v : $arg;
Branch: 1 branch
BranchTrueFalseTotal
defined $v ? : 1 0 1
31 }
32 }
33  
34 # Automatically compute the ranks if given data.
35 15 ▶ if( $self->x_data && $self->y_data &&
Branch: 1 branch
BranchTrueFalseTotal
if ($self->x_data and $self->y_data and @{$self->x_data;} and @{$self->y_data;}) 14 1 15
Condition66%6/9
$self->x_data and $self->y_data and @{$self->x_data;} and @{$self->y_data;}
!ll&&!rl&&r
1014
$self->x_data and $self->y_data and @{$self->x_data;}
!ll&&!rl&&r
1014
$self->x_data and $self->y_data
!ll&&!rl&&r
1014
MC/DC25%1/4
$self->x_data and $self->y_data and @{$self->x_data;} and @{$self->y_data;}
$self->x_data $self->y_data @{$self->x_data;} @{$self->y_data;}
Truth table40%2/5
A and B and C and D
A= $self->x_dataB= $self->y_dataC= @{$self->x_data;}D= @{$self->y_data;}
ABCDresult
0XXX0
10XX0
110X0
11100
11111
36 14 @{ $self->x_data } && @{ $self->y_data }
37 ) {
38 # "Co-normalize" the vectors if they are of unequal size.
39 14 my( $x, $y ) = pad_vectors( $self->x_data, $self->y_data );
40  
41 # "Co-sort" the bivariate data set by the first one.
42 14 ▶ ( $x, $y ) = co_sort( $x, $y ) if $self->{sorted};
Branch: 1 branch
BranchTrueFalseTotal
if $self->{"sorted"} 1 13 14
43  
44 # Set the massaged data.
45 14 $self->x_data( $x );
46 14 $self->y_data( $y );
47  
48 # Set the size of the data vector.
49 14 $self->size( scalar @{ $self->x_data } );
50  
51 # Set the ranks and ties of the vectors.
52 14 ( $x, $y ) = rank( $self->x_data );
53 14 $self->x_rank( $x );
54 14 $self->x_ties( $y );
55 14 ( $x, $y ) = rank( $self->y_data );
56 14 $self->y_rank( $x );
57 14 $self->y_ties( $y );
58 }
59 }
60  
61 196 ▶sub size {
Subroutine size: called CC=2 SCAR=6.9
Pod documented
62 196 my $self = shift;
63 196 ▶ $self->{size} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 14 182 196
64 196 return $self->{size};
65 }
66  
67 130 ▶sub x_data {
Subroutine x_data: called CC=2 SCAR=6.9
Pod documented
68 130 my $self = shift;
69 130 ▶ $self->{x_data} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 28 102 130
70 130 return $self->{x_data};
71 }
72  
73 101 ▶sub y_data {
Subroutine y_data: called CC=2 SCAR=6.9
Pod documented
74 101 my $self = shift;
75 101 ▶ $self->{y_data} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 28 73 101
76 101 return $self->{y_data};
77 }
78  
79 18 ▶sub x_rank {
Subroutine x_rank: called CC=2 SCAR=6.9
Pod documented
80 18 my $self = shift;
81 18 ▶ $self->{x_rank} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 14 4 18
82 18 return $self->{x_rank};
83 }
84  
85 19 ▶sub y_rank {
Subroutine y_rank: called CC=2 SCAR=6.9
Pod documented
86 19 my $self = shift;
87 19 ▶ $self->{y_rank} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 14 5 19
88 19 return $self->{y_rank};
89 }
90  
91 46 ▶sub x_ties {
Subroutine x_ties: called CC=2 SCAR=6.9
Pod documented
92 46 my $self = shift;
93 46 ▶ $self->{x_ties} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 14 32 46
94 46 return $self->{x_ties};
95 }
96  
97 44 ▶sub y_ties {
Subroutine y_ties: called CC=2 SCAR=6.9
Pod documented
98 44 my $self = shift;
99 44 ▶ $self->{y_ties} = shift if @_;
Branch: 1 branch
BranchTrueFalseTotal
if @_ 14 30 44
100 44 return $self->{y_ties};
101 }
102  
103 14 ▶sub spearman {
Subroutine spearman: called CC=6 SCAR=17.9
Pod documented
104 14 my $self = shift;
105 # Algorithm contributed by Jon Schutz <Jon.Schutz@youramigo.com>:
106 14 my($x_sum, $y_sum) = (0, 0);
107 14 $x_sum += $_ for @{$self->{x_rank}};
108 14 $y_sum += $_ for @{$self->{y_rank}};
109 14 my $n = $self->size;
110 14 my $x_mean = $x_sum / $n;
111 14 my $y_mean = $y_sum / $n;
112 # Compute the sum of the difference of the squared ranks.
113 14 my($x_sum2, $y_sum2, $xy_sum) = (0, 0, 0);
114 14 for( 0 .. $self->size - 1 ) {
115 119 $x_sum2 += ($self->{x_rank}[$_] - $x_mean) ** 2;
116 119 $y_sum2 += ($self->{y_rank}[$_] - $y_mean) ** 2;
117 119 $xy_sum += ($self->{x_rank}[$_] - $x_mean) * ($self->{y_rank}[$_] - $y_mean);
118 }
119 14 ▶ return 1 if $x_sum2 == 0 || $y_sum2 == 0;
Branch: 1 branch
BranchTrueFalseTotal
if $x_sum2 == 0 or $y_sum2 == 0 1 13 14
Condition66%2/3
$x_sum2 == 0 or $y_sum2 == 0
l!l&&r!l&&!r
1013
MC/DC50%1/2
$x_sum2 == 0 or $y_sum2 == 0
$x_sum2 == 0 $y_sum2 == 0
Truth table66%2/3
A or B
A= $x_sum2 == 0B= $y_sum2 == 0
ABresult
1X1
011
000
120 13 return $xy_sum / sqrt($x_sum2 * $y_sum2);
121 }
122  
123  
124 28 ▶sub rank {
Subroutine rank: called CC=6 SCAR=17.9
Pod documented
125 28 my $u = shift;
126  
127 # Make a list of tied ranks for each datum.
128 28 my %ties;
129 28 push @{ $ties{ $u->[$_] } }, $_ for 0 .. @$u - 1;
130  
131 28 my ($old, $cur) = (0, 0);
132  
133 # Set the averaged ranks.
134 28 my @ranks;
135 28 for my $x (sort { $a <=> $b } keys %ties) {
136 # Get the number of ties.
137 197 my $ties = @{ $ties{$x} };
138 197 $cur += $ties;
139  
140 197 ▶ if ($ties > 1) {
Branch: 1 branch
BranchTrueFalseTotal
if ($ties > 1) { } 15 182 197
141 # Average the tied data.
142 15 my $average = $old + ($ties + 1) / 2;
143 15 $ranks[$_] = $average for @{ $ties{$x} };
144 }
145 else {
146 # Add the single rank to the list of ranks.
147 182 $ranks[ $ties{$x}[0] ] = $cur;
148 }
149  
150 197 $old = $cur;
151 }
152  
153 # Remove the non-tied ranks.
154 28 delete @ties{ grep @{ $ties{$_} } <= 1, keys %ties };
155  
156 # Return the ranks arrayref in a scalar context and include ties
157 # if called in a list context.
158 28 ▶ return wantarray ? (\@ranks, \%ties) : \@ranks;
Branch: 1 branch
BranchTrueFalseTotal
wantarray ? : 28 0 28
159 }
160  
161 1 ▶sub co_sort {
Subroutine co_sort: called CC=3 SCAR=11.1
Pod documented
162 1 my( $u, $v ) = @_;
163 1 ▶ return unless @$u == @$v;
Branch: 1 branch
BranchTrueFalseTotal
unless @$u == @$v 0 1 1
164 # Ye olde Schwartzian Transforme:
165 $v = [
166 10 map { $_->[1] }
167 23 ▶ sort { $a->[0] <=> $b->[0] || $a->[1] <=> $b->[1] }
Condition66%2/3
$Statistics::RankCorrelation::a->[0] <=> $Statistics::RankCorrelation::b->[0] || $Statistics::RankCorrelation::a->[1] <=> $Statistics::RankCorrelation::b->[1]
l!l&&r!l&&!r
2120
MC/DC0%0/2
$Statistics::RankCorrelation::a->[0] <=> $Statistics::RankCorrelation::b->[0] || $Statistics::RankCorrelation::a->[1] <=> $Statistics::RankCorrelation::b->[1]
$Statistics::RankCorrelation::a->[0] <=> $Statistics::RankCorrelation::b->[0] $Statistics::RankCorrelation::a->[1] <=> $Statistics::RankCorrelation::b->[1]
Truth table66%2/3
A || B
A= $Statistics::RankCorrelation::a->[0] <=> $Statistics::RankCorrelation::b->[0]B= $Statistics::RankCorrelation::a->[1] <=> $Statistics::RankCorrelation::b->[1]
ABresult
1X1
011
000
168 1 map { [ $u->[$_], $v->[$_] ] }
169 0 .. @$u - 1
170 ];
171 # Sort the independent vector last.
172 1 $u = [ sort { $a <=> $b } @$u ];
173 1 return $u, $v;
174 }
175  
176 3 ▶sub csim {
Subroutine csim: called CC=4 SCAR=13.9
Pod documented
177 3 my $self = shift;
178  
179 # Get the pitch matrices for each vector.
180 3 my $m1 = correlation_matrix($self->{x_data});
181 #warn map { "@$_\n" } @$m1;
182 3 my $m2 = correlation_matrix($self->{y_data});
183 #warn map { "@$_\n" } @$m2;
184  
185 # Compute the rank correlation.
186 3 my $k = 0;
187 3 for my $i (0 .. @$m1 - 1) {
188 16 for my $j (0 .. @$m1 - 1) {
189 96 ▶ $k++ if $m1->[$i][$j] == $m2->[$i][$j];
Branch: 1 branch
BranchTrueFalseTotal
if $m1->[$i][$j] == $m2->[$i][$j] 40 56 96
190 }
191 }
192  
193 # Return the rank correlation normalized by the number of rows in
194 # the pitch matrices.
195 3 return $k / (@$m1 * @$m1);
196 }
197  
198 14 ▶sub pad_vectors {
Subroutine pad_vectors: called CC=3 SCAR=13.3
Pod documented
199 14 my ($u, $v) = @_;
200  
201 14 ▶ if (@$u > @$v) {
Branch: 2 branches
BranchTrueFalseTotal
if (@$u > @$v) { } 0 14 14
elsif (@$u < @$v) { } 0 14 14
202 0 $v = [ @$v, (0) x (@$u - @$v) ];
203 }
204 elsif (@$u < @$v) {
205 0 $u = [ @$u, (0) x (@$v - @$u) ];
206 }
207  
208 14 return $u, $v;
209 }
210  
211 6 ▶sub correlation_matrix {
Subroutine correlation_matrix: called CC=4 SCAR=13.9
Pod documented
212 6 my $u = shift;
213 6 my $c;
214  
215 # Is a row value (i) lower than a column value (j)?
216 6 for my $i (0 .. @$u - 1) {
217 32 for my $j (0 .. @$u - 1) {
218 192 ▶ $c->[$i][$j] = $u->[$i] < $u->[$j] ? 1 : 0;
Branch: 1 branch
BranchTrueFalseTotal
$u->[$i] < $u->[$j] ? : 68 124 192
219 }
220 }
221  
222 6 return $c;
223 }
224  
225 13 ▶sub kendall {
Subroutine kendall: called CC=10 SCAR=23.0
Pod documented
226 13 my $self = shift;
227  
228 # Calculate number of concordant and discordant pairs.
229 13 my( $concordant, $discordant ) = ( 0, 0 );
230 13 for my $i ( 0 .. $self->size - 1 ) {
231 115 for my $j ( $i + 1 .. $self->size - 1 ) {
232 574 my $x_sign = sign( $self->{x_data}[$j] - $self->{x_data}[$i] );
233 574 my $y_sign = sign( $self->{y_data}[$j] - $self->{y_data}[$i] );
234 574 ▶ if (not($x_sign and $y_sign)) {}
Branch: 2 branches
BranchTrueFalseTotal
if (not $x_sign && $y_sign) { } 108 466 574
elsif ($x_sign == $y_sign) { } 288 178 466
Condition100%3/3
$x_sign && $y_sign
!ll&&!rl&&r
7533466
MC/DC100%2/2
$x_sign && $y_sign
$x_sign $y_sign
Truth table100%3/3
A && B
A= $x_signB= $y_sign
ABresult
0X0
100
111
235 288 elsif ($x_sign == $y_sign) { $concordant++ }
236 178 else { $discordant++ }
237 }
238 }
239  
240 # Set the indirect relationship.
241 13 my $d = $self->size * ($self->size - 1) / 2;
242 13 ▶ if( keys %{ $self->x_ties } || keys %{ $self->y_ties } ) {
Branch: 1 branch
BranchTrueFalseTotal
if (keys %{$self->x_ties;} or keys %{$self->y_ties;}) 5 8 13
Condition100%3/3
keys %{$self->x_ties;} or keys %{$self->y_ties;}
l!l&&r!l&&!r
238
MC/DC100%2/2
keys %{$self->x_ties;} or keys %{$self->y_ties;}
keys %{$self->x_ties;} keys %{$self->y_ties;}
Truth table100%3/3
A or B
A= keys %{$self->x_ties;}B= keys %{$self->y_ties;}
ABresult
1X1
011
000
243 5 my $x = 0;
244 5 $x += @$_ * (@$_ - 1) for values %{ $self->x_ties };
245 5 $x = $d - $x / 2;
246 5 my $y = 0;
247 5 $y += @$_ * (@$_ - 1) for values %{ $self->y_ties };
248 5 $y = $d - $y / 2;
249 5 $d = sqrt($x * $y);
250 }
251  
252 13 return ($concordant - $discordant) / $d;
253 }
254  
255 1148 ▶sub sign {
Subroutine sign: called CC=3 SCAR=11.0
Pod documented
256 1148 my $x = shift;
257 1148 ▶ return 0 if $x == 0;
Branch: 1 branch
BranchTrueFalseTotal
if $x == 0 124 1024 1148
258 1024 ▶ return $x > 0 ? 1 : -1;
Branch: 1 branch
BranchTrueFalseTotal
$x > 0 ? : 621 403 1024
259 }
260  
261 2 1;
262  
263 __END__
↑ Summary