File size: 7,961 Bytes
7885a28
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import pytest
from pytest import raises as assert_raises, warns as assert_warns

import numpy as np
from numpy.testing import assert_approx_equal, assert_allclose, assert_equal

from scipy.spatial.distance import cdist
from scipy import stats

class TestMGCErrorWarnings:
    """ Tests errors and warnings derived from MGC.
    """
    def test_error_notndarray(self):
        # raises error if x or y is not a ndarray
        x = np.arange(20)
        y = [5] * 20
        assert_raises(ValueError, stats.multiscale_graphcorr, x, y)
        assert_raises(ValueError, stats.multiscale_graphcorr, y, x)

    def test_error_shape(self):
        # raises error if number of samples different (n)
        x = np.arange(100).reshape(25, 4)
        y = x.reshape(10, 10)
        assert_raises(ValueError, stats.multiscale_graphcorr, x, y)

    def test_error_lowsamples(self):
        # raises error if samples are low (< 3)
        x = np.arange(3)
        y = np.arange(3)
        assert_raises(ValueError, stats.multiscale_graphcorr, x, y)

    def test_error_nans(self):
        # raises error if inputs contain NaNs
        x = np.arange(20, dtype=float)
        x[0] = np.nan
        assert_raises(ValueError, stats.multiscale_graphcorr, x, x)

        y = np.arange(20)
        assert_raises(ValueError, stats.multiscale_graphcorr, x, y)

    def test_error_wrongdisttype(self):
        # raises error if metric is not a function
        x = np.arange(20)
        compute_distance = 0
        assert_raises(ValueError, stats.multiscale_graphcorr, x, x,
                      compute_distance=compute_distance)

    @pytest.mark.parametrize("reps", [
        -1,    # reps is negative
        '1',   # reps is not integer
    ])
    def test_error_reps(self, reps):
        # raises error if reps is negative
        x = np.arange(20)
        assert_raises(ValueError, stats.multiscale_graphcorr, x, x, reps=reps)

    def test_warns_reps(self):
        # raises warning when reps is less than 1000
        x = np.arange(20)
        reps = 100
        assert_warns(RuntimeWarning, stats.multiscale_graphcorr, x, x, reps=reps)

    def test_error_infty(self):
        # raises error if input contains infinities
        x = np.arange(20)
        y = np.ones(20) * np.inf
        assert_raises(ValueError, stats.multiscale_graphcorr, x, y)


class TestMGCStat:
    """ Test validity of MGC test statistic
    """
    def _simulations(self, samps=100, dims=1, sim_type=""):
        # linear simulation
        if sim_type == "linear":
            x = np.random.uniform(-1, 1, size=(samps, 1))
            y = x + 0.3 * np.random.random_sample(size=(x.size, 1))

        # spiral simulation
        elif sim_type == "nonlinear":
            unif = np.array(np.random.uniform(0, 5, size=(samps, 1)))
            x = unif * np.cos(np.pi * unif)
            y = (unif * np.sin(np.pi * unif) +
                 0.4*np.random.random_sample(size=(x.size, 1)))

        # independence (tests type I simulation)
        elif sim_type == "independence":
            u = np.random.normal(0, 1, size=(samps, 1))
            v = np.random.normal(0, 1, size=(samps, 1))
            u_2 = np.random.binomial(1, p=0.5, size=(samps, 1))
            v_2 = np.random.binomial(1, p=0.5, size=(samps, 1))
            x = u/3 + 2*u_2 - 1
            y = v/3 + 2*v_2 - 1

        # raises error if not approved sim_type
        else:
            raise ValueError("sim_type must be linear, nonlinear, or "
                             "independence")

        # add dimensions of noise for higher dimensions
        if dims > 1:
            dims_noise = np.random.normal(0, 1, size=(samps, dims-1))
            x = np.concatenate((x, dims_noise), axis=1)

        return x, y

    @pytest.mark.xslow
    @pytest.mark.parametrize("sim_type, obs_stat, obs_pvalue", [
        ("linear", 0.97, 1/1000),           # test linear simulation
        ("nonlinear", 0.163, 1/1000),       # test spiral simulation
        ("independence", -0.0094, 0.78)     # test independence simulation
    ])
    def test_oned(self, sim_type, obs_stat, obs_pvalue):
        np.random.seed(12345678)

        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type=sim_type)

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y)
        assert_approx_equal(stat, obs_stat, significant=1)
        assert_approx_equal(pvalue, obs_pvalue, significant=1)

    @pytest.mark.xslow
    @pytest.mark.parametrize("sim_type, obs_stat, obs_pvalue", [
        ("linear", 0.184, 1/1000),           # test linear simulation
        ("nonlinear", 0.0190, 0.117),        # test spiral simulation
    ])
    def test_fived(self, sim_type, obs_stat, obs_pvalue):
        np.random.seed(12345678)

        # generate x and y
        x, y = self._simulations(samps=100, dims=5, sim_type=sim_type)

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y)
        assert_approx_equal(stat, obs_stat, significant=1)
        assert_approx_equal(pvalue, obs_pvalue, significant=1)

    @pytest.mark.xslow
    def test_twosamp(self):
        np.random.seed(12345678)

        # generate x and y
        x = np.random.binomial(100, 0.5, size=(100, 5))
        y = np.random.normal(0, 1, size=(80, 5))

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y)
        assert_approx_equal(stat, 1.0, significant=1)
        assert_approx_equal(pvalue, 0.001, significant=1)

        # generate x and y
        y = np.random.normal(0, 1, size=(100, 5))

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y, is_twosamp=True)
        assert_approx_equal(stat, 1.0, significant=1)
        assert_approx_equal(pvalue, 0.001, significant=1)

    @pytest.mark.xslow
    def test_workers(self):
        np.random.seed(12345678)

        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type="linear")

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y, workers=2)
        assert_approx_equal(stat, 0.97, significant=1)
        assert_approx_equal(pvalue, 0.001, significant=1)

    @pytest.mark.xslow
    def test_random_state(self):
        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type="linear")

        # test stat and pvalue
        stat, pvalue, _ = stats.multiscale_graphcorr(x, y, random_state=1)
        assert_approx_equal(stat, 0.97, significant=1)
        assert_approx_equal(pvalue, 0.001, significant=1)

    @pytest.mark.xslow
    def test_dist_perm(self):
        np.random.seed(12345678)
        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type="nonlinear")
        distx = cdist(x, x, metric="euclidean")
        disty = cdist(y, y, metric="euclidean")

        stat_dist, pvalue_dist, _ = stats.multiscale_graphcorr(distx, disty,
                                                               compute_distance=None,
                                                               random_state=1)
        assert_approx_equal(stat_dist, 0.163, significant=1)
        assert_approx_equal(pvalue_dist, 0.001, significant=1)

    @pytest.mark.fail_slow(20)  # all other tests are XSLOW; we need at least one to run
    @pytest.mark.slow
    def test_pvalue_literature(self):
        np.random.seed(12345678)

        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type="linear")

        # test stat and pvalue
        _, pvalue, _ = stats.multiscale_graphcorr(x, y, random_state=1)
        assert_allclose(pvalue, 1/1001)

    @pytest.mark.xslow
    def test_alias(self):
        np.random.seed(12345678)

        # generate x and y
        x, y = self._simulations(samps=100, dims=1, sim_type="linear")

        res = stats.multiscale_graphcorr(x, y, random_state=1)
        assert_equal(res.stat, res.statistic)