BUG/REF ? kwargs in RegressionModel.fit OLS, ...

[josef-pkt]

It looks like we can still have misspelled, incorrect keywords in fit of linear regression models.
https://stackoverflow.com/questions/75842157/results-summary-reporting-z-instead-of-t-value-for-multiple-ols-regression-using

Any kwargs are just added as attributes to the results instance.

I don't know or don't remember what this is used for, maybe some subclasses use it.

If subclasses use it, then we have to make up a list of allowed keywords similar to what I did for __init__ in models.

[josef-pkt]

The kwargs have been added 13 years ago, but I don't see a specific reason
b6eaa4c#diff-4038c404a4188bbadc3f50dd9db62c486e4806cacda9df176ea7e8a2b53388f0

[josef-pkt]

unit test fails if I raise a ValueError if kwargs in

FAILED ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_gretl.py::TestGLSARGretl::test_all
FAILED ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_gretl.py::test_GLSARlag
FAILED ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_regression.py::test_regularized_predict
ERROR ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_stata.py::TestGLSARCorc::test_params_table
ERROR ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_stata.py::TestGLSARCorc::test_predicted
ERROR ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_stata.py::TestGLSARCorc::test_rho
ERROR ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_stata.py::TestGLSARCorc::test_glsar_arima
ERROR ..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_glsar_stata.py::TestGLSARCorc::test_glsar_iter0

So we are using the kwargs.

example GLSAR.iterative_fit uses it to attach extra attributes to results instance

results = self.fit(history=history, **kwargs)

one possible solution is to add a more explicit results_kwds to fit

[josef-pkt]

for fit_regularized unit test

It looks like the generic code always uses maxiter as fit keyword.
maxiter does not apply or do anything in OLS, WLS, and GLSAR uses maxiter only in iterative_fit

..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\tests\test_regression.py:1468:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\linear_model.py:848: in fit_regularized
    rslt = OLS(self.wendog, self.wexog).fit_regularized(
..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\regression\linear_model.py:1130: in fit_regularized
    return fit_elasticnet(self, method=method,
..\eclipse-workspace\statsmodels_gh\statsmodels\statsmodels\base\elastic_net.py:236: in fit_elasticnet
    rslt = model1.fit(maxiter=0)