statsmodels.gam.generalized_additive_model.GLMGam.fit#

GLMGam.fit(start_params=None, maxiter=1000, method='pirls', tol=1e-08, scale=None, cov_type='nonrobust', cov_kwds=None, use_t=None, full_output=True, disp=False, max_start_irls=3, **kwargs)[source]#

Estimate parameters and create instance of GLMGamResults class

Most parameters are the same as for GLM and are only used to create the results instance.

Parameters:
start_paramsarray_like, optional

Initial guess of the solution for the loglikelihood maximization. If None, then the default for method="pirls" uses the family-specific starting mu, otherwise it is passed through to the underlying optimizer.

maxiterint, optional

Maximum number of iterations. Default is 1000.

methodstr

The special optimization method is “pirls” which uses a penalized version of IRLS. This is the default. Other methods are gradient optimizers as used in base.model.LikelihoodModel.fit that are called on the penalized log-likelihood.

tolfloat

Convergence tolerance for “pirls”. Default is 1e-8.

scalestr or float, optional

scale can be ‘X2’, ‘dev’, or a float. See GLM.fit for details.

cov_typestr

The type of parameter estimate covariance matrix to compute.

cov_kwdsdict-like

Extra arguments for calculating the covariance of the parameter estimates.

use_tbool

If True, the Student t-distribution is used for inference.

full_outputbool, optional

Set to True to have all available output in the Results object’s mle_retvals attribute. Not used if method is “pirls”.

dispbool, optional

Set to True to print convergence messages. Not used if method is “pirls”.

max_start_irlsint

The number of PIRLS iterations used to obtain starting values for gradient optimization. Only relevant if method is set to something other than “pirls”.

**kwargs

Additional keyword arguments used in the call to the underlying optimizer.

Returns:
resinstance of wrapped GLMGamResults