65 lines
2.2 KiB
Python
65 lines
2.2 KiB
Python
|
import pytest
|
||
|
|
import numpy as np
|
||
|
|
from numpy.testing import assert_allclose
|
||
|
|
import scipy.special._ufuncs as scu
|
||
|
|
from scipy.integrate import tanhsinh
|
||
|
|
|
||
|
|
|
||
|
|
type_char_to_type_tol = {'f': (np.float32, 32*np.finfo(np.float32).eps),
|
||
|
|
'd': (np.float64, 32*np.finfo(np.float64).eps)}
|
||
|
|
|
||
|
|
|
||
|
|
# Each item in this list is
|
||
|
|
# (func, args, expected_value)
|
||
|
|
# All the values can be represented exactly, even with np.float32.
|
||
|
|
#
|
||
|
|
# This is not an exhaustive test data set of all the functions!
|
||
|
|
# It is a spot check of several functions, primarily for
|
||
|
|
# checking that the different data types are handled correctly.
|
||
|
|
test_data = [
|
||
|
|
(scu._beta_pdf, (0.5, 2, 3), 1.5),
|
||
|
|
(scu._beta_pdf, (0, 1, 5), 5.0),
|
||
|
|
(scu._beta_pdf, (1, 5, 1), 5.0),
|
||
|
|
(scu._beta_ppf, (0.5, 5., 5.), 0.5), # gh-21303
|
||
|
|
(scu._binom_cdf, (1, 3, 0.5), 0.5),
|
||
|
|
(scu._binom_pmf, (1, 4, 0.5), 0.25),
|
||
|
|
(scu._hypergeom_cdf, (2, 3, 5, 6), 0.5),
|
||
|
|
(scu._nbinom_cdf, (1, 4, 0.25), 0.015625),
|
||
|
|
(scu._ncf_mean, (10, 12, 2.5), 1.5),
|
||
|
|
]
|
||
|
|
|
||
|
|
|
||
|
|
@pytest.mark.parametrize('func, args, expected', test_data)
|
||
|
|
def test_stats_boost_ufunc(func, args, expected):
|
||
|
|
type_sigs = func.types
|
||
|
|
type_chars = [sig.split('->')[-1] for sig in type_sigs]
|
||
|
|
for type_char in type_chars:
|
||
|
|
typ, rtol = type_char_to_type_tol[type_char]
|
||
|
|
args = [typ(arg) for arg in args]
|
||
|
|
# Harmless overflow warnings are a "feature" of some wrappers on some
|
||
|
|
# platforms. This test is about dtype and accuracy, so let's avoid false
|
||
|
|
# test failures cause by these warnings. See gh-17432.
|
||
|
|
with np.errstate(over='ignore'):
|
||
|
|
value = func(*args)
|
||
|
|
assert isinstance(value, typ)
|
||
|
|
assert_allclose(value, expected, rtol=rtol)
|
||
|
|
|
||
|
|
|
||
|
|
def test_landau():
|
||
|
|
# Test that Landau distribution ufuncs are wrapped as expected;
|
||
|
|
# accuracy is tested by Boost.
|
||
|
|
x = np.linspace(-3, 10, 10)
|
||
|
|
args = (0, 1)
|
||
|
|
res = tanhsinh(lambda x: scu._landau_pdf(x, *args), -np.inf, x)
|
||
|
|
cdf = scu._landau_cdf(x, *args)
|
||
|
|
assert_allclose(res.integral, cdf)
|
||
|
|
sf = scu._landau_sf(x, *args)
|
||
|
|
assert_allclose(sf, 1-cdf)
|
||
|
|
ppf = scu._landau_ppf(cdf, *args)
|
||
|
|
assert_allclose(ppf, x)
|
||
|
|
isf = scu._landau_isf(sf, *args)
|
||
|
|
assert_allclose(isf, x, rtol=1e-6)
|
||
|
|
|
||
|
|
def test_gh22956():
|
||
|
|
_ = scu._ncx2_pdf(30, 1e307, 16)
|