sklearn compliance and default subestimator

[Peter9192]

In checking sklearn compliance, it complains about the fixed-effects model as input, even though this very similar to the documented example (under get_params).

Not sure if this is a bad idea, or whether the checks are too strict. See scikit-learn/scikit-learn#28257

An alternative could be to have a dynamic class generator. Something like this:

def create_mixed_effects_model(Model):
    """Create a mixed-effects model with Model as fixed-effects model."""
    class_name = "MixedEffects" + Model.__name__
    class_dict = dict(fe_model = Model)
    new_class = type(class_name, (MixedEffectsModel,), class_dict)
    return new_class

However, this would not be in line with the documented example for get_params.