Fepois predict with newdata

[leostimpfle]

Closes #468

[leostimpfle]

predict still does not work with newdata and fixed effects (see #467). Something doesn't quite work in the fixef method inherited from feols. I'll investigate further.

[codecov]

Codecov Report

Attention: Patch coverage is 0% with 6 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
pyfixest/estimation/fepois_.py	0.00%	6 Missing ⚠️

Flag	Coverage Δ
core-tests	77.79% <0.00%> (+0.10%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
pyfixest/estimation/fepois_.py	86.79% <0.00%> (-0.42%)	⬇️

[s3alfisc]

If I change to the following

    def predict(
        self,
        newdata: Optional[DataFrameType] = None,
        atol: float = 1e-6,
        btol: float = 1e-6,
        type: PredictionType = "link",
    ) -> np.ndarray:

        #if newdata is not None and self._has_fixef:
        #    raise NotImplementedError()

        y_hat = super().predict(newdata=newdata, type=type, atol=atol, btol=btol)
        if type == "response":
            y_hat = np.exp(y_hat)
        return y_hat

predict works with newdata and fixed effects for type = "response":

%load_ext autoreload
%autoreload 2

import pyfixest as pf
data = pf.get_data(model = "Fepois")

fit = pf.fepois("Y ~ X1 | f1", data = data)
fit.predict(newdata = data.dropna(), type = "response")[0:5]
# array([2.7803393 , 2.81678101, 3.20245421, 3.3940836 , 3.38641651])
fit.predict(type = "response")[0:5]
# array([2.78032513, 2.81699512, 3.20381613, 3.39393412, 3.38457835])