def null(self):
_endog = self._endog
wls = WLS(_endog, np.ones((len(_endog), 1)), weights=self._data_weights)
return wls.fit().fittedvalues
def fit(self, maxiter=100, method="IRLS", tol=1e-8, data_weights=1.0, scale=None):
"""
Fits a generalized linear model for a given family.
parameters
----------
data_weights : array-like or scalar, only used with Binomial
Number of trials for each observation. Used for only for
binomial data when `endog` is specified as a 2d array of
(successes, failures). Note that this argument will be
dropped in the future.
maxiter : int, optional
Default is 100.
method : string
Default is 'IRLS' for iteratively reweighted least squares. This
is currently the only method available for GLM fit.
scale : string or float, optional
`scale` can be 'X2', 'dev', or a float
The default value is None, which uses `X2` for Gamma, Gaussian,
and Inverse Gaussian.
`X2` is Pearson's chi-square divided by `df_resid`.
The default is 1 for the Binomial and Poisson families.
`dev` is the deviance divided by df_resid
tol : float
Convergence tolerance. Default is 1e-8.
"""
if np.shape(data_weights) != () and not isinstance(self.family, families.Binomial):
raise ValueError, "Data weights are only to be supplied for\
the Binomial family"
self.data_weights = data_weights
if np.shape(self.data_weights) == () and self.data_weights > 1:
self.data_weights = self.data_weights * np.ones((self.exog.shape[0]))
self.scaletype = scale
if isinstance(self.family, families.Binomial):
# thisc checks what kind of data is given for Binomial. family will need a reference to
# endog if this is to be removed from the preprocessing
self.endog = self.family.initialize(self.endog)
mu = self.family.starting_mu(self.endog)
wlsexog = self.exog
eta = self.family.predict(mu)
self.iteration += 1
dev = self.family.deviance(self.endog, mu)
if np.isnan(dev):
raise ValueError, "The first guess on the deviance function \
returned a nan. This could be a boundary problem and should be reported."
else:
self.history["deviance"].append(dev)
# first guess on the deviance is assumed to be scaled by 1.
while (
np.fabs(self.history["deviance"][self.iteration] - self.history["deviance"][self.iteration - 1])
) > tol and self.iteration < maxiter:
self.weights = data_weights * self.family.weights(mu)
wlsendog = eta + self.family.link.deriv(mu) * (self.endog - mu)
# - offset
wls_results = WLS(wlsendog, wlsexog, self.weights).fit()
eta = np.dot(self.exog, wls_results.params) # + offset
mu = self.family.fitted(eta)
self._update_history(wls_results, mu)
self.scale = self.estimate_scale(mu)
self.iteration += 1
self.mu = mu
glm_results = GLMResults(self, wls_results.params, wls_results.normalized_cov_params, self.scale)
glm_results.bse = np.sqrt(np.diag(wls_results.cov_params(scale=self.scale)))
return glm_results
