def test_two_averaged_spectra(self):
"""
For nspec=2, I can derive this by hand:
"""
power = 1.0
nspec = 2
manual_pval = 1.0 - avg_cdf_two_spectra(power)
assert np.isclose(cospectra_pvalue(power, nspec), manual_pval)
def test_single_spectrum_with_positive_power(self):
"""
Because the Laplace distribution is always symmetric
around zero, let's do a second version where I look
for a different number.
"""
power = 0.69314718055
nspec = 1
assert np.isclose(cospectra_pvalue(power, nspec), 0.25)
def test_single_spectrum(self):
# the Laplace distribution is symmetric around
# 0, so a power of 0 should return p=0.5
power = 0.0
nspec = 1
assert cospectra_pvalue(power, nspec) == 0.5
def test_pval_fails_if_nspec_not_integer(self):
power = 1.0
nspec = 1.5
with pytest.raises(ValueError):
pval = cospectra_pvalue(power, nspec)
def test_pval_fails_if_nspec_negative(self):
power = 1.0
nspec = -10
with pytest.raises(ValueError):
pval = cospectra_pvalue(power, nspec)
def test_pval_fails_if_multiple_powers_inf(self):
power = [1, 2.0, np.inf]
nspec = 1
with pytest.raises(ValueError):
pval = cospectra_pvalue(power, nspec)
def test_pval_fails_if_single_power_nan(self):
power = np.nan
nspec = 1
with pytest.raises(ValueError):
pval = cospectra_pvalue(power, nspec)
def test_pval_returns_iterable_when_iterable_input(self):
power = [0, 1, 2]
nspec = 1.0
pval = cospectra_pvalue(power, nspec)
assert isinstance(pval, np.ndarray)
assert len(pval) == len(power)
def test_pval_returns_float_when_float_input(self):
power = 1.0
nspec = 1.0
pval = cospectra_pvalue(power, nspec)
assert isinstance(pval, float)
def test_sixty_spectra(self):
power = 1.0
nspec = 60
gauss = scipy.stats.norm(0, np.sqrt(2 / (nspec + 1)))
pval_theory = gauss.sf(power)
assert np.isclose(cospectra_pvalue(power, nspec), pval_theory)