def test_influence(self):
# currently only smoke test
res1 = self.res1
from statsmodels.stats.outliers_influence import MLEInfluence
influ = MLEInfluence(res1)
attrs = [
'cooks_distance', 'd_fittedvalues', 'd_fittedvalues_scaled',
'd_params', 'dfbetas', 'hat_matrix_diag', 'resid_studentized'
]
for attr in attrs:
getattr(influ, attr)
influ.summary_frame()
def test_influence(self):
# currently only smoke test
res1 = self.res1
from statsmodels.stats.outliers_influence import MLEInfluence
influ0 = MLEInfluence(res1)
influ = res1.get_influence()
attrs = ['cooks_distance', 'd_fittedvalues', 'd_fittedvalues_scaled',
'd_params', 'dfbetas', 'hat_matrix_diag', 'resid_studentized'
]
for attr in attrs:
getattr(influ, attr)
frame = influ.summary_frame()
frame0 = influ0.summary_frame()
assert_allclose(frame, frame0, rtol=1e-13, atol=1e-13)
def setup_class(cls):
df = data_bin
res = GLM(df['constrict'], df[['const', 'log_rate', 'log_volumne']],
family=families.Binomial()).fit(attach_wls=True, atol=1e-10)
cls.infl1 = res.get_influence()
cls.infl0 = MLEInfluence(res)
def get_influence(self):
"""
Influence and outlier measures
See notes section for influence measures that do not apply for
zero inflated models.
Returns
-------
MLEInfluence
The instance has methods to calculate the main influence and
outlier measures as attributes.
See Also
--------
statsmodels.stats.outliers_influence.MLEInfluence
Notes
-----
ZeroInflated models have functions that are not differentiable
with respect to sample endog if endog=0. This means that generalized
leverage cannot be computed in the usual definition.
Currently, both the generalized leverage, in `hat_matrix_diag`
attribute and studetized residuals are not available. In the influence
plot generalized leverage is replaced by a hat matrix diagonal that
only takes combined exog into account, computed in the same way as
for OLS. This is a measure for exog outliers but does not take
specific features of the model into account.
"""
# same as sumper in DiscreteResults, only added for docstring
from statsmodels.stats.outliers_influence import MLEInfluence
return MLEInfluence(self)
def get_influence(self):
"""
Get an instance of MLEInfluence with influence and outlier measures
Returns
-------
infl : MLEInfluence instance
The instance has methods to calculate the main influence and
outlier measures as attributes.
See Also
--------
statsmodels.stats.outliers_influence.MLEInfluence
Notes
-----
Support for mutli-link and multi-exog models is still experimental
in MLEInfluence. Interface and some definitions might still change.
Note: Difference to R betareg: Betareg has the same general leverage
as this model. However, they use a linear approximation hat matrix
to scale and studentize influence and residual statistics.
MLEInfluence uses the generalized leverage as hat_matrix_diag.
Additionally, MLEInfluence uses pearson residuals for residual
analusis.
References
----------
todo
"""
from statsmodels.stats.outliers_influence import MLEInfluence
return MLEInfluence(self)
def setup_class(cls):
df = data_bin
mod = GLM(df['constrict'], df[['const', 'log_rate', 'log_volumne']],
family=families.Binomial(link=families.links.probit()))
res = mod.fit(method="newton", tol=1e-10)
from statsmodels.discrete.discrete_model import Probit
mod2 = Probit(df['constrict'], df[['const', 'log_rate', 'log_volumne']])
res2 = mod2.fit(method="newton", tol=1e-10)
cls.infl1 = MLEInfluence(res) # res.get_influence()
cls.infl0 = res2.get_influence()
def setup_class(cls):
from .test_diagnostic import get_duncan_data
endog, exog, labels = get_duncan_data()
data = pd.DataFrame(np.column_stack((endog, exog)),
columns='y const var1 var2'.split(),
index=labels)
res = GLM.from_formula('y ~ const + var1 + var2 - 1', data).fit()
#res = GLM(endog, exog).fit()
cls.infl1 = res.get_influence()
cls.infl0 = MLEInfluence(res)
def setup_class(cls):
yi = np.array([0, 2, 14, 19, 30])
ni = 40 * np.ones(len(yi))
xi = np.arange(1, len(yi) + 1)
exog = np.column_stack((np.ones(len(yi)), xi))
endog = np.column_stack((yi, ni - yi))
res = GLM(endog, exog, family=families.Binomial()).fit()
cls.infl1 = res.get_influence()
cls.infl0 = MLEInfluence(res)
cls.cd_rtol = 5e-5