def left_side(a, b, c, d):
num_steps = min(a, d)
p_val = []
for i in range(num_steps):
a -= 1
b += 1
c += 1
d -= 1
p_val.append(_hypergeom_distribution(a, b, c, d))
return p_val
def right_side(a, b, c, d):
num_steps = min(b, c)
p_val = []
for i in range(num_steps):
a += 1
b -= 1
c -= 1
d += 1
p_val.append(_hypergeom_distribution(a, b, c, d))
return p_val
def test_hypergeom_Dict_Error(self):
a, b, c, d = 10, 10, 10, {'d': 10}
with pytest.raises(TypeError):
utils._hypergeom_distribution(a, b, c, d)
def test_hypergeom_Float_Error(self):
a, b, c, d = 10, 10, 10, 10.5
with pytest.raises(TypeError):
utils._hypergeom_distribution(a, b, c, d)
def test_hypergeom_String_Error(self):
a, b, c, d = 10, 10, 10, '10'
with pytest.raises(TypeError):
utils._hypergeom_distribution(a, b, c, d)
def test_hypergeom_Result(self):
a, b, c, d = 1, 2, 3, 4
assert pytest.approx(0.5, 0.01) == utils._hypergeom_distribution(
a, b, c, d)
def fisher_test(cont_table, alternative='two-sided'):
"""Found in scipy.stats as fisher_exact
Used to determine the exact likelihood that we would observe a measurement in our 2x2 contingency table that
is just as extreme, if not moreso, than our observed results.
Parameters
----------
cont_table: list or numpy array, 2 x 2
A 2x2 contingency table
alternative: str, {two-sided, greater, less}, default is two-sided
Our alternative hypothesis
Return
------
p: float, 0 <= p <= 1
The exact likelihood of finding a more extreme measurement than our observed data
"""
if alternative.casefold() not in ['two-sided', 'greater', 'less']:
raise ValueError("Cannot determine method for alternative hypothesis")
cont_table = _check_table(cont_table, True)
if cont_table.shape != (2, 2):
raise AttributeError(
"Fisher's Exact Test is meant for a 2x2 contingency table")
a, b, c, d = cont_table[0, 0], cont_table[0,
1], cont_table[1,
0], cont_table[1,
1]
p = _hypergeom_distribution(a, b, c, d)
# left side
def left_side(a, b, c, d):
num_steps = min(a, d)
p_val = []
for i in range(num_steps):
a -= 1
b += 1
c += 1
d -= 1
p_val.append(_hypergeom_distribution(a, b, c, d))
return p_val
# right side
def right_side(a, b, c, d):
num_steps = min(b, c)
p_val = []
for i in range(num_steps):
a += 1
b -= 1
c -= 1
d += 1
p_val.append(_hypergeom_distribution(a, b, c, d))
return p_val
if alternative.casefold() == 'greater':
right_p_val = right_side(a, b, c, d)
return p + np.sum(right_p_val)
elif alternative.casefold() == 'less':
left_p_val = left_side(a, b, c, d)
return p + np.sum(left_p_val)
else:
left_p_val, right_p_val = left_side(a, b, c, d), right_side(a, b, c, d)
all_p = right_p_val + left_p_val
return p + np.sum([i for i in all_p if i <= p])