def test(self):
_, expected_pvalue = spearmanr(x, y)
test_pvalue_nospeedup = stats.compute_permutation_test_pvalue(
x,
y,
pdist_metric='spearman',
permute_func='ecdf',
speedup=False,
iters=1000,
seed=123)
self.assertLessEqual(abs(test_pvalue_nospeedup - expected_pvalue), eps)
def test_compute_permutation_test_pvalue_insignificant_speedup(self):
np.random.seed(2)
x = np.random.normal(size=1000)
y = np.random.normal(size=1000)
test_pvalue = stats.compute_permutation_test_pvalue(
x,
y,
pdist_metric='pearson',
permute_func='ecdf',
speedup=True,
alpha=0.05,
iters=1000,
seed=123)
# we don't care about the exact value
self.assertTrue(test_pvalue > 0.05)
def test_compute_permutation_test_pvalue_insignificant(self):
np.random.seed(2)
eps = 0.02
x = np.random.normal(size=1000)
y = np.random.normal(size=1000)
_, expected_pvalue = pearsonr(x, y)
test_pvalue = stats.compute_permutation_test_pvalue(
x,
y,
pdist_metric='pearson',
permute_func='ecdf',
speedup=False,
iters=1000,
seed=123)
self.assertLessEqual(abs(test_pvalue - expected_pvalue), eps)
def test(self):
_, expected_pvalue = spearmanr(x, y)
test_pvalue = stats.compute_permutation_test_pvalue(
x,
y,
pdist_metric='spearman',
permute_func='gpd',
speedup=True,
alpha=0.05,
iters=1000,
seed=123)
if expected_pvalue <= 0.05: # significant
self.assertLessEqual(abs(test_pvalue - expected_pvalue), eps)
else:
self.assertTrue(test_pvalue > 0.05)
def test_compute_permutation_test_pvalue_significant(self):
np.random.seed(1)
eps = 0.001
# source: https://machinelearningmastery.com/how-to-calculate-nonparametric-rank-correlation-in-python/
x = np.random.normal(size=1000) * 20
y = x + np.random.normal(size=1000) * 10
# get expected p-value
_, expected_pvalue = spearmanr(x, y)
test_pvalue = stats.compute_permutation_test_pvalue(
x,
y,
pdist_metric='spearman',
permute_func='ecdf',
speedup=False,
iters=1000,
seed=123)
self.assertLessEqual(abs(test_pvalue - expected_pvalue), eps)