def test_assoc(stat, expected):
# 2d Array
obs1 = np.array([[12, 13, 14, 15, 16],
[17, 16, 18, 19, 11],
[9, 15, 14, 12, 11]])
a = association(observed=obs1, method=stat)
assert_allclose(a, expected)
def fill_pairwisematrix(self):
'''
fills the square matrix using scipy's association method
Returns
-------
None.
'''
n = len(self.cat_columns)
# get all possible pair-wise combinations in the columns list
# this assumes that A-->B equals B-->A so we don't need to
# calculate the same thing twice
# we also never get "A --> A"
all_combinations = combinations(self.cat_columns, r=2)
# note that because we ignore redundant combinations,
# we perform half the calculations, so we get the results
# twice as fast
for comb in all_combinations:
i = comb[0]
j = comb[1]
# make contingency table
input_tab = crosstab(self.data[i], self.data[j])
# find the resulting cramer association value using scipy's
# association method
res_cramer = association(input_tab, method='cramer')
self.matrix[i][j], self.matrix[j][i] = res_cramer, res_cramer