-
Notifications
You must be signed in to change notification settings - Fork 267
/
base.py
276 lines (217 loc) · 9.41 KB
/
base.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
#! /usr/bin/env python
# ----------------------------------------------------------------------------
# Copyright (c) 2013--, scikit-bio development team.
#
# Distributed under the terms of the Modified BSD License.
#
# The full license is in the file COPYING.txt, distributed with this software.
# ----------------------------------------------------------------------------
from __future__ import absolute_import, division
from future.utils.six import StringIO
import csv
import numpy as np
import pandas as pd
from skbio.core.distance import DistanceMatrix
class CategoricalStats(object):
"""Base class for categorical statistical methods.
Categorical statistical methods generally test for significant differences
between discrete groups of objects, as determined by a categorical variable
(grouping vector).
See Also
--------
ANOSIM, PERMANOVA
"""
short_method_name = ''
long_method_name = ''
test_statistic_name = ''
def __init__(self, distance_matrix, grouping, column=None):
if not isinstance(distance_matrix, DistanceMatrix):
raise TypeError("Input must be a DistanceMatrix.")
if isinstance(grouping, pd.DataFrame):
if column is None:
raise ValueError("Must provide a column name if supplying a "
"data frame.")
else:
grouping = self._df_to_vector(distance_matrix, grouping,
column)
elif column is not None:
raise ValueError("Must provide a data frame if supplying a column "
"name.")
if len(grouping) != distance_matrix.shape[0]:
raise ValueError("Grouping vector size must match the number of "
"IDs in the distance matrix.")
# Find the group labels and convert grouping to an integer vector
# (factor).
groups, grouping = np.unique(grouping, return_inverse=True)
if len(groups) == len(grouping):
raise ValueError("All values in the grouping vector are unique. "
"This method cannot operate on a grouping vector "
"with only unique values (e.g., there are no "
"'within' distances because each group of "
"objects contains only a single object).")
if len(groups) == 1:
raise ValueError("All values in the grouping vector are the same. "
"This method cannot operate on a grouping vector "
"with only a single group of objects (e.g., "
"there are no 'between' distances because there "
"is only a single group).")
self._dm = distance_matrix
self._grouping = grouping
self._groups = groups
self._tri_idxs = np.triu_indices(self._dm.shape[0], k=1)
def _df_to_vector(self, distance_matrix, df, column):
"""Return a grouping vector from a data frame column.
Parameters
----------
distance_marix : DistanceMatrix
Distance matrix whose IDs will be mapped to group labels.
df : pandas.DataFrame
Data frame (indexed by distance matrix ID).
column : str
Column name in `df` containing group labels.
Returns
-------
list
Grouping vector (vector of labels) based on the IDs in
`distance_matrix`. Each ID's label is looked up in the data frame
under the column specified by `column`.
Raises
------
ValueError
If `column` is not in the data frame, or a distance matrix ID is
not in the data frame.
"""
if column not in df:
raise ValueError("Column '%s' not in data frame." % column)
grouping = df.loc[distance_matrix.ids, column]
if grouping.isnull().any():
raise ValueError("One or more IDs in the distance matrix are not "
"in the data frame.")
return grouping.tolist()
def __call__(self, permutations=999):
"""Execute the statistical method.
Parameters
----------
permutations : int, optional
Number of permutations to use when calculating statistical
significance. Must be >= 0. If 0, the resulting p-value will be
``None``.
Returns
-------
CategoricalStatsResults
Results of the method, including test statistic and p-value.
.. shownumpydoc
"""
if permutations < 0:
raise ValueError("Number of permutations must be greater than or "
"equal to zero.")
stat = self._run(self._grouping)
p_value = None
if permutations > 0:
perm_stats = np.empty(permutations, dtype=np.float64)
for i in range(permutations):
perm_grouping = np.random.permutation(self._grouping)
perm_stats[i] = self._run(perm_grouping)
p_value = ((perm_stats >= stat).sum() + 1) / (permutations + 1)
return CategoricalStatsResults(self.short_method_name,
self.long_method_name,
self.test_statistic_name,
self._dm.shape[0], self._groups, stat,
p_value, permutations)
def _run(self, grouping):
raise NotImplementedError("Subclasses must implement _run().")
class CategoricalStatsResults(object):
"""Statistical method results container.
Stores the results of running a `CategoricalStats` method a single time,
and provides a way to format the results.
Attributes
----------
short_method_name
long_method_name
test_statistic_name
sample_size
groups
statistic
p_value
permutations
Notes
-----
Users will generally not directly instantiate objects of this class. The
various categorical statistical methods will return an object of this type
when they are run.
"""
def __init__(self, short_method_name, long_method_name,
test_statistic_name, sample_size, groups, statistic, p_value,
permutations):
self.short_method_name = short_method_name
self.long_method_name = long_method_name
self.test_statistic_name = test_statistic_name
self.sample_size = sample_size
self.groups = groups
self.statistic = statistic
self.p_value = p_value
self.permutations = permutations
def __str__(self):
"""Return pretty-print (fixed width) string."""
rows = (self._format_header(), self._format_data())
max_widths = []
for col_idx in range(len(rows[0])):
max_widths.append(max(map(lambda e: len(e[col_idx]), rows)))
results = []
for row in rows:
padded_row = []
for col_idx, val in enumerate(row):
padded_row.append(val.rjust(max_widths[col_idx]))
results.append(' '.join(padded_row))
return '\n'.join(results) + '\n'
def _repr_html_(self):
"""Return a string containing an HTML table of results.
This method will be called within the IPython Notebook instead of
__repr__ to display results.
"""
header = self._format_header()
data = self._format_data()
return pd.DataFrame([data[1:]], columns=header[1:],
index=[data[0]])._repr_html_()
def summary(self, delimiter='\t'):
"""Return a formatted summary of results as a string.
The string is formatted as delimited text.
Parameters
----------
delimiter : str, optional
String to delimit fields by in formatted output. Default is tab
(TSV).
Returns
-------
str
Delimited-text summary of results.
"""
summary = StringIO()
csv_writer = csv.writer(summary, delimiter=delimiter,
lineterminator='\n')
csv_writer.writerow(self._format_header())
csv_writer.writerow(self._format_data())
return summary.getvalue()
def _format_header(self):
return ('Method name', 'Sample size', 'Number of groups',
self.test_statistic_name, 'p-value', 'Number of permutations')
def _format_data(self):
p_value_str = self._format_p_value(self.p_value, self.permutations)
return (self.short_method_name, '%d' % self.sample_size,
'%d' % len(self.groups), str(self.statistic), p_value_str,
'%d' % self.permutations)
def _format_p_value(self, p_value, permutations):
"""Format p-value as a string with the correct number of decimals.
Number of decimals is determined by the number of permutations.
"""
if p_value is None:
result = 'N/A'
elif permutations < 10:
# This can be the last step of a long process, so we don't want to
# fail.
result = ('Too few permutations to compute p-value (permutations '
'= %d)' % permutations)
else:
decimal_places = int(np.log10(permutations + 1))
result = ('%1.' + '%df' % decimal_places) % p_value
return result