A simple wrapper class to estimate a Rare Event Logit model of King and Zeng (2001) in Python.

• This module is provided 'as is' and is prone to errors.

– In a terminal window install the requirements as:

pip install -r requirements.txt

– In Python environment import the relogit module as:

from relogit import relogit

– Specify the function using the following variables:

Y : array_like
            A 1-d endogenous response variable. See statsmodels guidance.
            
X : array_like
            A nobs x k array where nobs is the number of observations and k is 
            the number of regressors. An intercept is added by setting add_const
            to True.
            
add_const : Boolean, optional
            Whether to add a constant into X. The default is False.
            
disp : Boolean, optional
            Whether to display details for fitting. The default is False.
            See statsmodels guidance  

– Train a RE-Logit model by

relogit_model=relogit(Y, X, *optional keywords*)

– Get estimations of unbiased probability predicted_relogit and unbiased coefficients coeffs_unbiased by RE-Logit for a new set of exogenous set X_test as:

predicted_relogit,coeffs_unbiased = relogit_model.predict(X_test)

– Get additional estimations of probability predicted_logit and coefficients coeff_biased by the Logit for the same input X_test as:

predicted_relogit,coeffs_unbiased,predicted_logit,coeff_biased = relogit_model.predict(X_test)

– For more see the accompanying example script vignette.py

The following packages are required to use this module:

• Numpy
• statsmodels