Output Fama-Macbeth Academic Results

Download the script

import pandas as pd

import numpy as np

from linearmodels import FamaMacBeth

from decimal import Decimal

def fm_single_summary(fm_single_df):

    '''

    This function is an auxiliary of the core function `fm_summary`. It inputs a dataframe of a single fama-macbeth regression.\n

    The columns are all the variables' names, the indexes are 'parameter' and 'tstat'.\n

    It returns a pd.Series of the single fama-macbeth regression.

    '''

    df = fm_single_df.copy()

    var_list = []

    ## keep 2 decimals of the t-stat in parentheses, like (1.96)

    tstat_list = ['(' + str(Decimal("%.2f" % float(i))) +')' if np.isnan(i)==False else '' for i in df.loc['tstat', :] ]

    df.loc['tstat', :] = tstat_list

    for variable in df.columns:

        for j in ['parameter', 'tstat']:

            var_list.append(df.loc[j, variable])

    summ = pd.Series(var_list)

    index_list = []

    for m in range(len(summ)):

        if (m % 2) == 0:

            index_list.append(df.columns[int(m/2)])

        elif (m % 2) !=0:

            index_list.append('')

    summ.index = index_list

    return summ

def fm_summary(reg_list):

    """

    :param `reg_list`: the list of regressions that need to be combined and summarized.\n

    :return: the summary of all the regressions in `reg_list`, which  conforms to academic norms.

    """

    total_df = pd.DataFrame()

    r2_list = []

    for reg in reg_list:

        reg_df = pd.DataFrame([reg.params, reg.tstats])

        total_df = pd.concat([total_df, reg_df])

        try:

            r2_list.append(reg.avg_adj_rsquared)

        except:

            r2_list.append(reg.rsquared)

    fm_summ_df = []

    for i in range(len(reg_list)):

        df_block = total_df.iloc[int(i*2):int(i*2+2),:]

        fm_summ = fm_single_summary(df_block) # for each reg, get its single summary

        fm_summ_df.append(fm_summ) # then combined all the summaries

    total_summ = pd.concat(fm_summ_df, axis=1)

    total_summ = total_summ.fillna('') # fill the nan with empty term

    ## the column names are positive integers starting from 1

    total_summ.columns = [str(i+1) for i in range(len(reg_list))]

    ## the next 4 rows are for getting the 'Avg. R Square', 'Total Obs.', 'Avg. Obs.'

    ## can be adjusted if more statistics are needed

    r2_list = pd.Series(r2_list, name='Avg. R Square', index=total_summ.columns)

    total_obs_list = pd.Series([reg.nobs for reg in reg_list], name='Total Obs.', index=total_summ.columns)

    mean_obs_list = pd.Series([reg.time_info['mean'] for reg in reg_list], name='Avg. Obs.', index=total_summ.columns)

    total_summ = total_summ.append([r2_list, total_obs_list, mean_obs_list])

    return total_summ

print(fm_summary([reg, reg1, reg2]))