def test_mixed_conditions(self):
self.assertTrue(
series_are_equivalent(self.p_AB__C__D_1,
condition(given(self.joint, D=1), 'C')))
self.assertTrue(
series_are_equivalent(self.p_AB__C_2__D,
condition(given(self.joint, C=2), 'D')))
def test_given_conditions(self):
self.assertTrue(
series_are_equivalent(self.p_ABC__D_1, given(self.joint, D=1)))
for c in ['A', 'B', 'C']:
kwargs = {c: 1}
self.assertFalse(
series_are_equivalent(self.p_ABC__D_1,
given(self.joint, **kwargs)))
self.assertTrue(
series_are_equivalent(self.p_AB__C_1__D_2,
given(self.joint, C=1, D=2)))
for c1, c2 in product(self.vars, self.vars):
if c1 == c2 or (c1 == 'C' and c2 == 'D'):
continue
kwargs = {c1: 1, c2: 2}
self.assertFalse(
series_are_equivalent(self.p_AB__C_1__D_2,
given(self.joint, **kwargs)))
self.assertTrue(
series_are_equivalent(self.p_A__B_1__C_2__D_3,
given(self.joint, B=1, C=2, D=3)))
def given(self, **given_vals) -> DiscreteDistribution:
"""
Return the Discrete Distribution at the given values of the conditional variables.
:param given_vals: Values of conditional variables to create probability distribution from.
"""
# check input arguments
given_val_keys = set(given_vals.keys())
has_all_conds = set(
self._cond_vars).intersection(given_val_keys) == set(
self._cond_vars)
if not has_all_conds:
raise ValueError(
'Must supply values for all conditioned variables to get to joint distribution.'
)
# calculate probability
data = given(self._data, **given_vals)
return DiscreteDistribution(data, given_conditions=given_vals)
def given(self, **given_conditions) -> 'DiscreteDistribution':
"""
Condition on values of variables.
:param given_conditions: Dict[{name}__{comparator}, value] for each conditioned variable.
"""
# check input variables
names_comps = set(given_conditions.keys())
if not all([valid_name_comparator(name_comp, self._joints) for name_comp in names_comps]):
raise ValueError('Given variables must be members of joint distribution.')
# calculate conditional distribution
data = given(self._data, **given_conditions)
cond_values = {**self._given_conditions, **given_conditions}
return DiscreteDistribution(
data=data,
given_var_names=self._given_vars + [cond_name(k, self.var_names) for k in names_comps],
given_conditions=cond_values
)
def test_chain_mixed_conditions(self):
self.assertFalse(
series_are_equivalent(condition(given(self.joint, C=1), 'D'),
given(condition(self.joint, 'D'), C=1)))
def test_chain_given_conditions(self):
self.assertTrue(
series_are_equivalent(given(given(self.joint, C=1), D=2),
given(given(self.joint, D=2), C=1)))