Pyfixest

Breaking API changes

It is no longer required to initiate an object of type `Fixest` prior to running `feols` or `fepois`. Instead,
you can now simply use `feols()` and `fepois()` as functions, just as in `fixest`. Both function can be found in an
`estimation` module and need to obtain a `pd.DataFrame` as a function argument:

py
from pyfixest.estimation import fixest, fepois
from pyfixest.utils import get_data

data = get_data()
fit = feols("Y ~ X1 | f1", data = data, vcov = "iid")


Calling `feols()` will return an instance of class `Feols`, while calling `fepois()` will return an instance of class `Fepois`.
Multiple estimation syntax will return an instance of class `FixestMulti`.

Post processing works as before via `.summary()`, `.tidy()` and other methods.

New Features

A summary function allows to compare multiple models:

py
from pyfixest.summarize import summary
fit2 = feols("Y ~ X1 + X2| f1", data = data, vcov = "iid")
summary([fit, fit2])


Visualization is possible via custom methods (`.iplot()` & `.coefplot()`), but a new module allows to visualize
a list of `Feols` and/or `Fepois` instances:

py
from pyfixest.visualize import coefplot, iplot
coefplot([fit, fit2])


The documentation has been improved (though there is still room for progress), and the code has been cleaned up a
bit (also lots of room for improvements).

This release ...

- provides more stringent unit testing
- fixes a few bugs (most importantly, for separation tests with multiple fixed effects)
- reimplements the IWLS Poisson algorithm (it now more closely follows naming conventions from the `pplmhdfe` paper)
- drops `Fepois` special `get_vcov` methods, it is now integrated into `Feols.get_vcov`
- installs `PyHDFE` directly from PyPi (as `PyHDFE` 0.2 now supports weights)
- bases the convergence criterion of the Poisson on it's absolute deviance. The implementation now matches `fixest` and `glm`

New Features

This release adds ...

- support for Poisson regression (via the algorithm suggested by Correia et al and implemented in their `pplmhdfe` package)
- a `predict()` method
- a `fixef()` method to retrieve fixed effects of estimated models (only for OLS)
- some goodness-of-fit measures (only for OLS)

For example, you can now run a Poisson Regression as

python
import pyfixest as pf
from pyfixest.utils import get_poisson_data

pdata = get_poisson_data()
fixest = pf.Fixest(data = pdata)
fixest.fepois("Y~X1", vcov = "hetero")

fixest.summary()


Model: Poisson
Dep. var.: Y
Inference: hetero
Observations: 1000

Estimate Std. Error t value Pr(>|t|) ci_l ci_u
6.035479 0.080534 74.942848 0.0 5.877635 6.193324
0.771267 0.071402 10.801754 0.0 0.631322 0.911213
---

- Fixes a bug in the computation of confidence intervals that lead to wildly incorrect CIs.
- Introduces a `confint()` method for `Fixest` objects.

- adds a `Fixest.fetch_model()` method to retrieve individual models out of Fixest
- adds post estimation routines for objects of type Feols: coef(), se(), confint,tidy(), etc

After estimating multiple models, it is now possible to fetch individual models (objects of type `Feols`) from an object of type `Fixest` via the `fetch_model()` method:

py
from pyfixest import Fixest
from pyfixest.utils import get_data
import pandas as pd
data = get_data()

fixest = Fixest(data)
fixest.feols('Y ~ X1 | csw(X2,X3)', vcov = 'iid')

mod = fixest.fetch_model(1)
Model: Y~X1|X2+X3
mod
<pyfixest.feols.Feols at 0x2da61c41b10>

mod.tidy()

Estimate Std. Error t value Pr(>|t|) confint_lower confint_upper
coefnames
X1 0.03975 0.120714 0.329288 0.741973 -0.03218 0.047319

- adds support for Python 3.11
- makes wildboottest.py optional in order to do so (wildboottest depends on numba, which does not support python 3.12)