Currently, Poisson regression does not support regression weights, mostly because I have not thought through the statistics of regression weights for iterated weighted least squares algorithms.
To do
Research the statistics of iterated weighted least squares algorithms with regression weights.
Implement weights and fweights argument for the Fepois class.
Unlock the weights argument for the fepois function (it is currently blocked).
Drop the tests for ValueError for useage of regression weights for fepois.
Context
Currently, Poisson regression does not support regression weights, mostly because I have not thought through the statistics of regression weights for iterated weighted least squares algorithms.
To do
weights
andfweights
argument for the Fepois class.fepois
function (it is currently blocked).fepois
.fixest
in tests/test_vs_fixest.py::test_single_fit.