line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
1
|
|
|
|
|
|
|
package Dumbbench::Stats; |
2
|
3
|
|
|
3
|
|
17
|
use strict; |
|
3
|
|
|
|
|
6
|
|
|
3
|
|
|
|
|
71
|
|
3
|
3
|
|
|
3
|
|
12
|
use warnings; |
|
3
|
|
|
|
|
5
|
|
|
3
|
|
|
|
|
57
|
|
4
|
3
|
|
|
3
|
|
12
|
use List::Util (); |
|
3
|
|
|
|
|
5
|
|
|
3
|
|
|
|
|
33
|
|
5
|
3
|
|
|
3
|
|
1228
|
use Statistics::CaseResampling (); |
|
3
|
|
|
|
|
1562
|
|
|
3
|
|
|
|
|
99
|
|
6
|
|
|
|
|
|
|
|
7
|
|
|
|
|
|
|
use Class::XSAccessor { |
8
|
3
|
|
|
|
|
20
|
constructor => 'new', |
9
|
|
|
|
|
|
|
accessors => [qw/data name/], |
10
|
3
|
|
|
3
|
|
20
|
}; |
|
3
|
|
|
|
|
4
|
|
11
|
|
|
|
|
|
|
|
12
|
|
|
|
|
|
|
# Note: This is entirely unoptimized. There is a lot of unnecessary |
13
|
|
|
|
|
|
|
# stuff going on. This is to allow the user to modify the data |
14
|
|
|
|
|
|
|
# set in flight. If this comes back to haunt us at some point, |
15
|
|
|
|
|
|
|
# we can still optimize, but at this point, convenience still wins. |
16
|
|
|
|
|
|
|
|
17
|
|
|
|
|
|
|
sub sorted_data { |
18
|
0
|
|
|
0
|
0
|
0
|
my $self = shift; |
19
|
0
|
|
|
|
|
0
|
my $sorted = [sort { $a <=> $b } @{$self->data}]; |
|
0
|
|
|
|
|
0
|
|
|
0
|
|
|
|
|
0
|
|
20
|
0
|
|
|
|
|
0
|
return $sorted; |
21
|
|
|
|
|
|
|
} |
22
|
|
|
|
|
|
|
|
23
|
2
|
|
|
2
|
0
|
14
|
sub first_quartile { Statistics::CaseResampling::first_quartile($_[0]->data) } |
24
|
2
|
|
|
2
|
0
|
470
|
sub second_quartile { return $_[0]->median } |
25
|
2
|
|
|
2
|
0
|
18
|
sub third_quartile { Statistics::CaseResampling::third_quartile($_[0]->data) } |
26
|
|
|
|
|
|
|
|
27
|
|
|
|
|
|
|
|
28
|
16
|
|
|
16
|
0
|
23
|
sub n { scalar(@{$_[0]->data}) } |
|
16
|
|
|
|
|
51
|
|
29
|
|
|
|
|
|
|
|
30
|
|
|
|
|
|
|
sub sum { |
31
|
16
|
|
|
16
|
0
|
19
|
my $self = shift; |
32
|
16
|
|
|
|
|
30
|
return List::Util::sum(@{$self->data}); |
|
16
|
|
|
|
|
71
|
|
33
|
|
|
|
|
|
|
} |
34
|
|
|
|
|
|
|
|
35
|
|
|
|
|
|
|
sub min { |
36
|
0
|
|
|
0
|
0
|
0
|
my $self = shift; |
37
|
0
|
|
|
|
|
0
|
return List::Util::min(@{$self->data}); |
|
0
|
|
|
|
|
0
|
|
38
|
|
|
|
|
|
|
} |
39
|
|
|
|
|
|
|
|
40
|
|
|
|
|
|
|
sub max { |
41
|
0
|
|
|
0
|
0
|
0
|
my $self = shift; |
42
|
0
|
|
|
|
|
0
|
return List::Util::max(@{$self->data}); |
|
0
|
|
|
|
|
0
|
|
43
|
|
|
|
|
|
|
} |
44
|
|
|
|
|
|
|
|
45
|
|
|
|
|
|
|
sub mean { |
46
|
16
|
|
|
16
|
0
|
1429
|
my $self = shift; |
47
|
16
|
|
|
|
|
34
|
return $self->sum / $self->n; |
48
|
|
|
|
|
|
|
} |
49
|
|
|
|
|
|
|
|
50
|
68
|
|
|
68
|
0
|
255
|
sub median { Statistics::CaseResampling::median($_[0]->data) } # O(n)! |
51
|
|
|
|
|
|
|
|
52
|
|
|
|
|
|
|
sub median_confidence_limits { |
53
|
0
|
|
|
0
|
0
|
0
|
my $self = shift; |
54
|
0
|
|
|
|
|
0
|
my $nsigma = shift; |
55
|
0
|
|
|
|
|
0
|
my $alpha = Statistics::CaseResampling::nsigma_to_alpha($nsigma); |
56
|
|
|
|
|
|
|
# note: The 1000 here is kind of a lower limit for reasonable accuracy. |
57
|
|
|
|
|
|
|
# But if the data set is small, that's more significant. If the data |
58
|
|
|
|
|
|
|
# set is VERY large, then running much more than 1k resamplings |
59
|
|
|
|
|
|
|
# is VERY expensive. So 1k is probably a reasonable default. |
60
|
0
|
|
|
|
|
0
|
return Statistics::CaseResampling::median_simple_confidence_limits($self->data, 1-$alpha, 1000) |
61
|
|
|
|
|
|
|
} |
62
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
sub mad { |
64
|
24
|
|
|
24
|
0
|
47
|
my $self = shift; |
65
|
24
|
|
|
|
|
41
|
my $median = $self->median; |
66
|
24
|
|
|
|
|
37
|
my @val = map {abs($_ - $median)} @{$self->data}; |
|
103
|
|
|
|
|
190
|
|
|
24
|
|
|
|
|
48
|
|
67
|
24
|
|
|
|
|
89
|
return ref($self)->new(data => \@val)->median; |
68
|
|
|
|
|
|
|
} |
69
|
|
|
|
|
|
|
|
70
|
|
|
|
|
|
|
sub mad_dev { |
71
|
1
|
|
|
1
|
0
|
3
|
my $self = shift; |
72
|
1
|
|
|
|
|
3
|
return $self->mad()*1.4826; |
73
|
|
|
|
|
|
|
} |
74
|
|
|
|
|
|
|
|
75
|
|
|
|
|
|
|
sub std_dev { |
76
|
2
|
|
|
2
|
0
|
20
|
my $self = shift; |
77
|
2
|
|
|
|
|
6
|
my $data = $self->data; |
78
|
2
|
|
|
|
|
4
|
my $mean = $self->mean; |
79
|
2
|
|
|
|
|
6
|
my $var = 0; |
80
|
2
|
|
|
|
|
10
|
$var += ($_-$mean)**2 for @$data; |
81
|
2
|
|
|
|
|
4
|
$var /= @$data - 1; |
82
|
2
|
|
|
|
|
6
|
return sqrt($var); |
83
|
|
|
|
|
|
|
} |
84
|
|
|
|
|
|
|
|
85
|
|
|
|
|
|
|
sub filter_outliers { |
86
|
13
|
|
|
13
|
0
|
1002
|
my $self = shift; |
87
|
13
|
|
|
|
|
37
|
my %opt = @_; |
88
|
13
|
|
100
|
|
|
44
|
my $var_measure = $opt{variability_measure} || 'mad'; |
89
|
13
|
|
|
|
|
21
|
my $n_sigma = $opt{nsigma_outliers}; |
90
|
|
|
|
|
|
|
|
91
|
|
|
|
|
|
|
# If outlier rejection is turned off... |
92
|
13
|
100
|
|
|
|
41
|
if (not $n_sigma) { |
|
|
50
|
|
|
|
|
|
93
|
1
|
|
|
|
|
7
|
return ($self->data, []); |
94
|
|
|
|
|
|
|
} |
95
|
|
|
|
|
|
|
elsif ($n_sigma < 0) { |
96
|
0
|
|
|
|
|
0
|
Carp::croak("A negative value for the number of 'sigmas' makes no sense"); |
97
|
|
|
|
|
|
|
} |
98
|
|
|
|
|
|
|
|
99
|
12
|
|
|
|
|
28
|
my $data = $self->data; |
100
|
|
|
|
|
|
|
|
101
|
12
|
|
|
|
|
23
|
my $median = $self->median; |
102
|
12
|
|
|
|
|
27
|
my $variability = $self->$var_measure; |
103
|
12
|
|
|
|
|
31
|
my @good; |
104
|
|
|
|
|
|
|
my @outliers; |
105
|
12
|
|
|
|
|
25
|
foreach my $x (@$data) { |
106
|
48
|
100
|
|
|
|
102
|
if (abs($x-$median) <= $variability*$n_sigma) { |
107
|
39
|
|
|
|
|
57
|
push @good, $x; |
108
|
|
|
|
|
|
|
} |
109
|
|
|
|
|
|
|
else { |
110
|
9
|
|
|
|
|
16
|
push @outliers, $x; |
111
|
|
|
|
|
|
|
} |
112
|
|
|
|
|
|
|
} |
113
|
|
|
|
|
|
|
|
114
|
12
|
|
|
|
|
41
|
return(\@good, \@outliers); |
115
|
|
|
|
|
|
|
} |
116
|
|
|
|
|
|
|
|
117
|
|
|
|
|
|
|
|
118
|
|
|
|
|
|
|
1; |