def test_long_chains_collider_bias_without_MI(
df_long_chains_and_collider_without_MI):
df = df_long_chains_and_collider_without_MI(size=10000)
indep = sci_is_independent(data=df, vars_1=['d'], vars_2=['b'])
assert indep == True
indep = sci_is_independent(data=df, vars_1=['b'], vars_2=['d'])
assert indep == True
def test_uniform_multinomial_with_4_possible_values_size_10000(
df_2_multinomial_indep_RVs):
params = {
"data": df_2_multinomial_indep_RVs(size=10000),
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": []
}
params["data"] = df_2_multinomial_indep_RVs(size=10000)
assert sci_is_independent(**params) == True
params["data"] = df_2_multinomial_indep_RVs(size=1000)
assert sci_is_independent(**params) == True
params["data"] = df_2_multinomial_indep_RVs(size=100)
assert sci_is_independent(**params) == True
def test_X_and_Y_cause_Z(df_X_and_Y_cause_Z):
params_1 = {
"data": df_X_and_Y_cause_Z(size=1000),
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": []
}
assert sci_is_independent(**params_1) == True
params_2 = {
"data": df_X_and_Y_cause_Z(size=1000),
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": ['z']
}
assert sci_is_independent(**params_2) == False
def test_X_causes_Y(df_X_causes_Y):
params = {
"data": df_X_causes_Y(size=100),
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": []
}
assert sci_is_independent(**params) == False
def test_X_and_Y_are_deterministic(df_X_and_Y_are_deterministic):
params = {
"data": df_X_and_Y_are_deterministic(size=100),
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": []
}
assert sci_is_independent(**params) == False
def test_3_multinom_RVs_MAR(df_Z_causes_X_Y_and_X_Z_causes_MI_Y):
size = 10000
df = df_Z_causes_X_Y_and_X_Z_causes_MI_Y(size=size)
params = {
"data": df,
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": ['z']
}
assert sci_is_independent(**params) == True
def test_X_Y_Z_indep(multinomial_RV):
size = 1000
df = pd.DataFrame({
'x': multinomial_RV(size=size),
'y': multinomial_RV(size=size),
'z': multinomial_RV(size=size),
})
params = {
"data": df,
"vars_1": ['x'],
"vars_2": ['y'],
"conditioning_set": ['z']
}
assert sci_is_independent(**params) == True
def test_spurious_edge(df_long_chains_and_collider_with_MI):
size = 10000
df = df_long_chains_and_collider_with_MI(size=size)
params = {
"data": df,
"vars_1": ['b'],
"vars_2": ['d'],
"conditioning_set": []
}
# we expect b-d to be a spurious edge because we're implicitly conditioning
# on the collider MI_b
assert sci_is_independent(**params) == False