def test_cv_unbiased_std():
for i in range(10):
values = [random.normalvariate(0, 1) for i in range(10)]
cv = ContinuousValue()
cv.update_batch(values)
assert (cv.unbiased_std() -
((len(values) * utils.std(values) /
(len(values) - 1)) / utils.c4(len(values))) < 0.00000000001)
def test_cv_unbiased_std():
for i in range(10):
values = [random.normalvariate(0, 1) for i in range(10)]
cv = ContinuousValue()
cv.update_batch(values)
assert (cv.unbiased_std() -
((len(values) * utils.std(values) / (len(values) - 1)) /
utils.c4(len(values))) < 0.00000000001)
def unbiased_std(self):
"""Returns an unbiased estimate of the std
This implementation uses Bessel's correction and Cochran's theorem:
`<https://en.wikipedia.org/wiki/Unbiased_estimation_of_standard_deviation#Bias_correction>`_
:return: an unbiased estimate of the std
:rtype: float
.. seealso:: :meth:`concept_formation.utils.c4`
"""
if self.num < 2:
return 0.0
return sqrt(self.meanSq / (self.num - 1)) / c4(self.num)
def unbiased_std(self):
"""
Returns an unbiased estimate of the std, but for n < 2 the std is
estimated to be 0.0.
This implementation uses Bessel's correction and Cochran's theorem:
`<https://en.wikipedia.org/wiki/Unbiased_estimation_of_standard_deviation#Bias_correction>`_
:return: an unbiased estimate of the std
:rtype: float
.. seealso:: :meth:`concept_formation.utils.c4`
"""
if self.num < 2:
return 0.0
return sqrt(self.meanSq / (self.num - 1)) / c4(self.num)
def test_c4():
with pytest.raises(ValueError):
utils.c4(1)
def test_c4():
with pytest.raises(ValueError):
utils.c4(1)