scipy 1.15.3__cp313-cp313t-win_amd64.whl → 1.16.0rc2__cp313-cp313t-win_amd64.whl

scipy/stats/tests/test_morestats.py

@@ -2,9 +2,11 @@
 #
 # Further enhancements and tests added by numerous SciPy developers.
 #
+import contextlib
 import math
-import warnings
+import re
 import sys
+import warnings
 from functools import partial
 
 import numpy as np
@@ -15,7 +17,7 @@ from numpy.testing import (assert_array_equal, assert_almost_equal,
                            suppress_warnings)
 import pytest
 from pytest import raises as assert_raises
-import re
+
 from scipy import optimize, stats, special
 from scipy.stats._morestats import _abw_state, _get_As_weibull, _Avals_weibull
 from .common_tests import check_named_results
@@ -25,8 +27,7 @@ from scipy.stats._distr_params import distcont
 from scipy.stats._axis_nan_policy import (SmallSampleWarning, too_small_nd_omit,
                                           too_small_1d_omit, too_small_1d_not_omit)
 
-from scipy.conftest import array_api_compatible
-from scipy._lib._array_api import array_namespace, is_numpy
+from scipy._lib._array_api import is_numpy
 from scipy._lib._array_api_no_0d import (
     xp_assert_close,
     xp_assert_equal,
@@ -365,6 +366,7 @@ class TestAnderson:
         with pytest.raises(ValueError, match=message):
             stats.anderson(x, 'weibull_min')
 
+    @pytest.mark.thread_unsafe
     def test_weibull_warning_error(self):
         # Check for warning message when there are too few observations
         # This is also an example in which an error occurs during fitting
@@ -502,6 +504,7 @@ class TestAndersonKSamp:
                                   tm[0:5], 4)
         assert_allclose(p, 0.0041, atol=0.00025)
 
+    @pytest.mark.thread_unsafe
     def test_R_kSamples(self):
         # test values generates with R package kSamples
         # package version 1.2-6 (2017-06-14)
@@ -738,7 +741,6 @@ class TestAnsari:
         assert_allclose(pval_l, 1-pval/2, atol=1e-12)
 
 
-@array_api_compatible
 class TestBartlett:
     def test_data(self, xp):
         # https://www.itl.nist.gov/div898/handbook/eda/section3/eda357.htm
@@ -760,10 +762,7 @@ class TestBartlett:
         attributes = ('statistic', 'pvalue')
         check_named_results(res, attributes, xp=xp)
 
-    @pytest.mark.skip_xp_backends(
-        "jax.numpy", cpu_only=True,
-        reason='`var` incorrect when `correction > n` (google/jax#21330)')
-    @pytest.mark.usefixtures("skip_xp_backends")
+    @pytest.mark.filterwarnings("ignore:invalid value encountered in divide")
     def test_empty_arg(self, xp):
         args = (g1, g2, g3, g4, g5, g6, g7, g8, g9, g10, [])
         args = [xp.asarray(arg) for arg in args]
@@ -1343,9 +1342,9 @@ class TestMood:
 
     def test_mood_alternative(self):
 
-        np.random.seed(0)
-        x = stats.norm.rvs(scale=0.75, size=100)
-        y = stats.norm.rvs(scale=1.25, size=100)
+        rng = np.random.RandomState(0)
+        x = stats.norm.rvs(scale=0.75, size=100, random_state=rng)
+        y = stats.norm.rvs(scale=1.25, size=100, random_state=rng)
 
         stat1, p1 = stats.mood(x, y, alternative='two-sided')
         stat2, p2 = stats.mood(x, y, alternative='less')
@@ -1420,6 +1419,7 @@ class TestProbplot:
         assert_allclose(osm1, osm2)
         assert_allclose(osr1, osr2)
 
+    @pytest.mark.thread_unsafe
     @pytest.mark.skipif(not have_matplotlib, reason="no matplotlib")
     def test_plot_kwarg(self):
         fig = plt.figure()
@@ -1457,10 +1457,12 @@ class TestProbplot:
                       (np.nan, np.nan, 0.0)))
 
     def test_array_of_size_one(self):
-        with np.errstate(invalid='ignore'):
+        message = "One or more sample arguments is too small..."
+        with (np.errstate(invalid='ignore'),
+              pytest.warns(SmallSampleWarning, match=message)):
             assert_equal(stats.probplot([1], fit=True),
                          ((np.array([0.]), np.array([1])),
-                          (np.nan, np.nan, 0.0)))
+                          (np.nan, np.nan, np.nan)))
 
 
 class TestWilcoxon:
@@ -1768,6 +1770,27 @@ class TestWilcoxon:
         res = stats.wilcoxon(np.zeros(5), method=method)
         assert_allclose(res, [0, 1])
 
+    def test_wilcoxon_axis_broadcasting_errors_gh22051(self):
+        # In previous versions of SciPy, `wilcoxon` gave an incorrect error
+        # message when `AxisError` was not found in the base NumPy namespace.
+        # Check that this is resolved with and without the ANP decorator.
+        message = "Array shapes are incompatible for broadcasting."
+        with pytest.raises(ValueError, match=message):
+            stats.wilcoxon([1, 2, 3], [4, 5])
+
+        message = "operands could not be broadcast together with..."
+        with pytest.raises(ValueError, match=message):
+            stats.wilcoxon([1, 2, 3], [4, 5], _no_deco=True)
+
+        AxisError = getattr(np, 'AxisError', None) or np.exceptions.AxisError
+        message = "source: axis 3 is out of bounds for array of dimension 1"
+        with pytest.raises(AxisError, match=message):
+            stats.wilcoxon([1, 2, 3], [4, 5, 6], axis=3)
+
+        message = "`axis` must be compatible with the shape..."
+        with pytest.raises(AxisError, match=message):
+            stats.wilcoxon([1, 2, 3], [4, 5, 6], axis=3, _no_deco=True)
+
 
 # data for k-statistics tests from
 # https://cran.r-project.org/web/packages/kStatistics/kStatistics.pdf
@@ -1777,12 +1800,11 @@ x_kstat = [16.34, 10.76, 11.84, 13.55, 15.85, 18.20, 7.51, 10.22, 12.52, 14.68,
            12.10, 15.02, 16.83, 16.98, 19.92, 9.47, 11.68, 13.41, 15.35, 19.11]
 
 
-@array_api_compatible
 class TestKstat:
     def test_moments_normal_distribution(self, xp):
-        np.random.seed(32149)
-        data = xp.asarray(np.random.randn(12345), dtype=xp.float64)
-        moments = xp.asarray([stats.kstat(data, n) for n in [1, 2, 3, 4]])
+        rng = np.random.RandomState(32149)
+        data = xp.asarray(rng.randn(12345), dtype=xp.float64)
+        moments = xp.stack([stats.kstat(data, n) for n in [1, 2, 3, 4]])
 
         expected = xp.asarray([0.011315, 1.017931, 0.05811052, 0.0754134],
                               dtype=data.dtype)
@@ -1792,8 +1814,9 @@ class TestKstat:
         m1 = stats.moment(data, order=1)
         m2 = stats.moment(data, order=2)
         m3 = stats.moment(data, order=3)
-        xp_assert_close(xp.asarray((m1, m2, m3)), expected[:-1], atol=0.02, rtol=1e-2)
+        xp_assert_close(xp.stack((m1, m2, m3)), expected[:-1], atol=0.02, rtol=1e-2)
 
+    @pytest.mark.filterwarnings("ignore:invalid value encountered in scalar divide")
     def test_empty_input(self, xp):
         if is_numpy(xp):
             with pytest.warns(SmallSampleWarning, match=too_small_1d_not_omit):
@@ -1805,7 +1828,7 @@ class TestKstat:
 
     def test_nan_input(self, xp):
         data = xp.arange(10.)
-        data = xp.where(data == 6, xp.asarray(xp.nan), data)
+        data = xp.where(data == 6, xp.nan, data)
 
         xp_assert_equal(stats.kstat(data), xp.asarray(xp.nan))
 
@@ -1835,8 +1858,8 @@ class TestKstat:
         xp_assert_close(res, xp.asarray(ref))
 
 
-@array_api_compatible
 class TestKstatVar:
+    @pytest.mark.filterwarnings("ignore:invalid value encountered in scalar divide")
     def test_empty_input(self, xp):
         x = xp.asarray([])
         if is_numpy(xp):
@@ -1849,14 +1872,13 @@ class TestKstatVar:
 
     def test_nan_input(self, xp):
         data = xp.arange(10.)
-        data = xp.where(data == 6, xp.asarray(xp.nan), data)
+        data = xp.where(data == 6, xp.nan, data)
 
         xp_assert_equal(stats.kstat(data), xp.asarray(xp.nan))
 
     @skip_xp_backends(np_only=True,
                       reason='input validation of `n` does not depend on backend')
-    @pytest.mark.usefixtures("skip_xp_backends")
-    def test_bad_arg(self):
+    def test_bad_arg(self, xp):
         # Raise ValueError is n is not 1 or 2.
         data = [1]
         n = 10
@@ -1974,9 +1996,6 @@ class TestPpccMax:
                             -0.71215366521264145, decimal=7)
 
 
-@skip_xp_backends('jax.numpy', reason="JAX arrays do not support item assignment")
-@pytest.mark.usefixtures("skip_xp_backends")
-@array_api_compatible
 class TestBoxcox_llf:
 
     @pytest.mark.parametrize("dtype", ["float32", "float64"])
@@ -2007,8 +2026,13 @@ class TestBoxcox_llf:
         llf2 = stats.boxcox_llf(lmbda, np.vstack([x, x]).T)
         xp_assert_close(xp.asarray([llf, llf]), xp.asarray(llf2), rtol=1e-12)
 
+    @pytest.mark.thread_unsafe
     def test_empty(self, xp):
-        assert xp.isnan(xp.asarray(stats.boxcox_llf(1, xp.asarray([]))))
+        message = "One or more sample arguments is too small..."
+        context = (pytest.warns(SmallSampleWarning, match=message) if is_numpy(xp)
+                   else contextlib.nullcontext())
+        with context:
+            assert xp.isnan(xp.asarray(stats.boxcox_llf(1, xp.asarray([]))))
 
     def test_gh_6873(self, xp):
         # Regression test for gh-6873.
@@ -2019,11 +2043,27 @@ class TestBoxcox_llf:
         xp_assert_close(llf, xp.asarray(-17.93934208579061))
 
     def test_instability_gh20021(self, xp):
-        data = xp.asarray([2003, 1950, 1997, 2000, 2009])
+        data = xp.asarray([2003, 1950, 1997, 2000, 2009], dtype=xp.float64)
         llf = stats.boxcox_llf(1e-8, data)
         # The expected value was computed with mpsci, set mpmath.mp.dps=100
         # expect float64 output for integer input
-        xp_assert_close(llf, xp.asarray(-15.32401272869016598, dtype=xp.float64))
+        xp_assert_close(llf, xp.asarray(-15.32401272869016598, dtype=xp.float64),
+                        rtol=1e-7)
+
+    def test_axis(self, xp):
+        data = xp.asarray([[100, 200], [300, 400]])
+        llf_axis_0 = stats.boxcox_llf(1, data, axis=0)
+        llf_0 = xp.asarray([
+            stats.boxcox_llf(1, data[:, 0]),
+            stats.boxcox_llf(1, data[:, 1]),
+        ])
+        xp_assert_close(llf_axis_0, llf_0)
+        llf_axis_1 = stats.boxcox_llf(1, data, axis=1)
+        llf_1 = xp.asarray([
+            stats.boxcox_llf(1, data[0, :]),
+            stats.boxcox_llf(1, data[1, :]),
+        ])
+        xp_assert_close(llf_axis_1, llf_1)
 
 
 # This is the data from GitHub user Qukaiyi, given as an example
@@ -2490,8 +2530,8 @@ class TestYeojohnson:
             return x_inv
 
         n_samples = 20000
-        np.random.seed(1234567)
-        x = np.random.normal(loc=0, scale=1, size=(n_samples))
+        rng = np.random.RandomState(1234567)
+        x = rng.normal(loc=0, scale=1, size=(n_samples))
 
         x_inv = _inverse_transform(x, lmbda)
         xt, maxlog = stats.yeojohnson(x_inv)
@@ -2633,7 +2673,6 @@ class TestYeojohnsonNormmax:
         assert np.allclose(lmbda, 1.305, atol=1e-3)
 
 
-@array_api_compatible
 class TestCircFuncs:
     # In gh-5747, the R package `circular` was used to calculate reference
     # values for the circular variance, e.g.:
@@ -2658,16 +2697,14 @@ class TestCircFuncs:
         M2 = stats.circmean(x, high=360)
         xp_assert_close(M2, M1, rtol=1e-5)
 
-        # plain torch var/std ddof=1, so we need array_api_compat torch
-        xp_test = array_namespace(x)
-        V1 = xp_test.var(x*xp.pi/180, correction=0)
+        V1 = xp.var(x*xp.pi/180, correction=0)
         # for small variations, circvar is approximately half the
         # linear variance
         V1 = V1 / 2.
         V2 = stats.circvar(x, high=360)
         xp_assert_close(V2, V1, rtol=1e-4)
 
-        S1 = xp_test.std(x, correction=0)
+        S1 = xp.std(x, correction=0)
         S2 = stats.circstd(x, high=360)
         xp_assert_close(S2, S1, rtol=1e-4)
 
@@ -2731,7 +2768,6 @@ class TestCircFuncs:
         x = xp.asarray([355, 5, 2, 359, 10, 350, np.nan])
         xp_assert_equal(test_func(x, high=360), xp.asarray(xp.nan))
 
-    @skip_xp_backends('cupy', reason='cupy/cupy#8391')
     @pytest.mark.parametrize("test_func,expected",
                              [(stats.circmean,
                                {None: np.nan, 0: 355.66582264, 1: 0.28725053}),
@@ -2790,7 +2826,7 @@ class TestCircFuncs:
         # "white-box" sanity check that no undue loss of precision is
         # introduced by conversion between (high - low) and (2 * pi).
 
-        x = xp.linspace(1e-9, 1e-8, 100)
+        x = xp.linspace(1e-9, 6e-9, 50)
         assert xp.all(xp.sin(x) == x) and xp.all(xp.cos(x) == 1.0)
 
         m = (x * (2 * xp.pi) / (2 * xp.pi)) != x
@@ -3024,7 +3060,6 @@ class TestMedianTest:
         res = stats.median_test(x, y, correction=correction)
         assert_equal((res.statistic, res.pvalue, res.median, res.table), res)
 
-@array_api_compatible
 class TestDirectionalStats:
     # Reference implementations are not available
     def test_directional_stats_correctness(self, xp):
@@ -3038,7 +3073,7 @@ class TestDirectionalStats:
                          np.cos(incl) * np.sin(decl),
                          np.sin(incl)),
                         axis=1)
-        
+
         decl = xp.asarray(decl.tolist())
         incl = xp.asarray(incl.tolist())
         data = xp.asarray(data.tolist())
@@ -3070,9 +3105,7 @@ class TestDirectionalStats:
                                    axis=1)
         dirstats = stats.directional_stats(testdata_vector)
         directional_mean = dirstats.mean_direction
-        xp_test = array_namespace(directional_mean)  # np needs atan2
-        directional_mean_angle = xp_test.atan2(directional_mean[1],
-                                               directional_mean[0])
+        directional_mean_angle = xp.atan2(directional_mean[1], directional_mean[0])
         directional_mean_angle = directional_mean_angle % (2 * xp.pi)
         circmean = stats.circmean(testdata)
         xp_assert_close(directional_mean_angle, circmean)
@@ -3197,7 +3230,6 @@ class TestFDRControl:
         assert_array_equal(stats.false_discovery_control([]), [])
 
 
-@array_api_compatible
 class TestCommonAxis:
     # More thorough testing of `axis` in `test_axis_nan_policy`,
     # but those tests aren't run with array API yet. This class
@@ -3215,11 +3247,11 @@ class TestCommonAxis:
         x = xp.asarray(rng.random((6, 7)))
 
         res = fun(x, **kwargs, axis=0)
-        ref = xp.asarray([fun(x[:, i], **kwargs) for i in range(x.shape[1])])
+        ref = xp.stack([fun(x[:, i], **kwargs) for i in range(x.shape[1])])
         xp_assert_close(res, ref)
 
         res = fun(x, **kwargs, axis=1)
-        ref = xp.asarray([fun(x[i, :], **kwargs) for i in range(x.shape[0])])
+        ref = xp.stack([fun(x[i, :], **kwargs) for i in range(x.shape[0])])
         xp_assert_close(res, ref)
 
         res = fun(x, **kwargs, axis=None)

scipy/stats/tests/test_mstats_basic.py

@@ -1567,9 +1567,9 @@ class TestCompareWithStats:
     def generate_xy_sample(self, n):
         # This routine generates numpy arrays and corresponding masked arrays
         # with the same data, but additional masked values
-        np.random.seed(1234567)
-        x = np.random.randn(n)
-        y = x + np.random.randn(n)
+        rng = np.random.RandomState(1234567)
+        x = rng.randn(n)
+        y = x + rng.randn(n)
         xm = np.full(len(x) + 5, 1e16)
         ym = np.full(len(y) + 5, 1e16)
         xm[0:len(x)] = x