def combine(self, p_values, ref_p_values=None):
"""
Takes a list of p_values and combines them into a single p_value by choosing the p_value of the first feature.
:param p_values: a list of p_values of features.
:param ref_p_values: p-values of reference observations (if needed).
:return: The combined p-value.
"""
p_values = check_p_values(p_values)
# check that the given position of the feature is valid
if self.feature_position < 0 or self.feature_position >= p_values.shape[
1]:
raise ValueError(
'The given feature position is invalid! The position should be in [0, %i)'
% len(p_values))
# # assert p_values is a list
# assert type(p_values) == list
#
# # check that p_values is not empty
# if len(p_values) == 0:
# raise ValueError('The given list of p_values is empty')
#
# # check that all elements in p_values are floats
# if not all(isinstance(x, (int, float)) for x in p_values):
# raise ValueError('The elements in p_values should all be of the type \'float\'')
combined_p_values = np.take(p_values, self.feature_position, axis=1)
return combined_p_values
def combine(self, p_values, ref_p_values=None):
"""
Takes a list of p_values and combines them into a single p_value by calculating the average.
:param p_values: a list of p_values of features.
:param ref_p_values: p-values of reference observations (if needed).
:return: The combined p-value.
"""
p_values = check_p_values(p_values)
# if type(p_values) == list:
# p_values = np.array([p_values])
# assert p_values is a list
# assert type(p_values) == list
# check that p_values is not empty
# if len(p_values) == 0:
# raise ValueError('The given list of p_values is empty')
# check that all elements in p_values are floats
# if not all(isinstance(x, (int, float)) for x in p_values):
# raise ValueError('The elements in p_values should all be of the type \'float\'')
combined_p_values = np.mean(p_values, axis=1)
return combined_p_values
def combine(self, p_values, ref_p_values=None):
"""
Takes a list of p_values and combines them into a single p_value using the Stouffer’s Z-score method.
:param p_values: a list of p_values of features.
:param ref_p_values: p-values of reference observations (if needed).
:return: The combined p-value.
"""
p_values = check_p_values(p_values)
# # assert p_values is a list
# assert type(p_values) == list
#
# # check that p_values is not empty
# if len(p_values) == 0:
# raise ValueError('The given list of p_values is empty')
#
# # check that all elements in p_values are floats
# if not all(isinstance(x, (int, float)) for x in p_values):
# raise ValueError('The elements in p_values should all be of the type \'float\'')
combined_p_values = np.apply_along_axis(lambda x: combine_pvalues(x, method='stouffer')[1], axis=1,
arr=p_values)
return combined_p_values
def combine(self, p_values, ref_p_values=None):
"""
Takes a list of p_values and combines them into a single p_value by calculating the average.
:param p_values: a list of p_values of features.
:param ref_p_values: p-values of reference observations.
:return: The combined p-value.
"""
### INPUT VALIDATION
p_values = check_p_values(p_values)
# # assert p_values is a list
# assert type(p_values) == list
#
# # check that p_values is not empty
# if len(p_values) == 0:
# raise ValueError('The given list of p_values is empty')
#
# # check that all elements in p_values are floats
# if not all(isinstance(x, (int, float)) for x in p_values):
# raise ValueError('The elements in p_values should all be of the type \'float\'')
# check that ref_p_values are given
if ref_p_values is None:
raise ValueError('The empirical combiner needs reference p values to calculate the combined p value!')
# assert ref_p_values is a ndarray
assert type(ref_p_values) == np.ndarray
# check that the reference p_values of each reference observation are all floats (or integers)
for ref in ref_p_values:
if not all(isinstance(x, (int, float)) for x in ref):
raise ValueError('The p_values of each reference observation should all be of the type \'float\'')
### FUNCTION CODE
min_ref_p_values = ref_p_values.min(axis=1)
combined_p_values = np.apply_along_axis(
lambda x: self.empirical(min(x), min_ref_p_values, weights=np.ones(len(min_ref_p_values)),
direction=self.direction), axis=1, arr=p_values)
return combined_p_values