Regression Module

The regression module contains a function for fixed effects regression similar to Stata’s areg function.

regression

finance_byu.regression.areg(formula, data=None, absorb=None, cluster=None)

Performs efficient fixed effects regression.

Parameters:

formula: string

Patsy / R style formula string. Variables in the formula string must be contained in data.

data: pandas.core.frame.DataFrame

Pandas DataFrame containing data for regression.

absorb: string

Categorical variable to be absorbed (must be in data)

cluster: string

Cluster variable (must be in data)

Returns:

output: statsmodels.regression.linear_model.RegressionResultsWrapper

A statsmodels regression results object.

Example

>>> from finance_byu.regression import areg
>>> import numpy as np
>>> import pandas as pd
>>>
>>> n_obs = 1.0e2
>>>
>>> df = pd.DataFrame(np.random.random((int(n_obs),4)))
>>> df = df.rename(columns={0:'ret',1:'beta',2:'logme',3:'logbm'})
>>> df['caldt'] = np.random.choice(10,n_obs)
>>> reg = areg('ret ~ beta + logme + logbm',data=df,absorb='caldt',cluster='caldt')
>>> reg.summary()

                                OLS Regression Results
==============================================================================
Dep. Variable:                    ret   R-squared:                       0.027
Model:                            OLS   Adj. R-squared:                 -0.107
Method:                 Least Squares   F-statistic:                     2.136
Date:                Sat, 14 Dec 2019   Prob (F-statistic):              0.166
Time:                        14:01:32   Log-Likelihood:                -7.8543
No. Observations:                 100   AIC:                             23.71
Df Residuals:                      87   BIC:                             34.13
Df Model:                           3
Covariance Type:              cluster
==============================================================================
                 coef    std err          z      P>|z|      [0.025      0.975]
------------------------------------------------------------------------------
Intercept      0.5400      0.089      6.095      0.000       0.366       0.714
beta           0.0480      0.045      1.062      0.288      -0.041       0.137
logme         -0.1484      0.074     -2.013      0.044      -0.293      -0.004
logbm          0.0243      0.121      0.200      0.841      -0.213       0.262
==============================================================================
Omnibus:                       24.428   Durbin-Watson:                   2.069
Prob(Omnibus):                  0.000   Jarque-Bera (JB):                5.333
Skew:                           0.033   Prob(JB):                       0.0695
Kurtosis:                       1.871   Cond. No.                         6.82
==============================================================================

Warnings:
[1] Standard Errors are robust to cluster correlation (cluster)