Python statsmodels robust cov_type='hac-panel' issue

问题

I hope this is the right place for my question.

I would like to understand how to use the 'hac-panel' cov_type when running sm.OLS. I have struggled with it the whole day but still cannot figure it out. Here is an example of my code (with data):

import pandas as pd
import statsmodels.api as sm
import statsmodels.formula.api as smf
from pandas.tseries.offsets import *

# Just grabbing some random data here
dat = sm.datasets.macrodata.load_pandas().data
dat['time'] = dat['year'].apply(lambda x: pd.to_datetime(x, format='%Y'))
dat['time'] = dat.apply(lambda x:(x['time'] + BQuarterBegin(x['quarter'])), axis=1)
dat = dat.set_index('time')
dat = dat.sort_index()
dat['dGDP'] = (dat['realgdp'] - dat['realgdp'].shift(1))/dat['realgdp'].shift(1) * 100.0
dat['dM1'] = (dat['m1'] - dat['m1'].shift(1))/dat['m1'].shift(1) * 100.0
dat['dUEMP'] = dat['unemp'] - dat['unemp'].shift(1)
dat['dCPI'] = dat['infl'] - dat['infl'].shift(1)
dat = dat[['dGDP', 'dM1', 'dUEMP', 'dCPI']]

# Fitting the model
y_var = dat.unstack()
x_var = pd.DataFrame(dat.shift(1).unstack(), columns=['01m']).combine_first(pd.DataFrame(dat.shift(3).unstack(), columns=['03m'])).combine_first(pd.DataFrame(dat.shift(12).unstack(), columns=['12m']))

model = sm.OLS(y_var, sm.add_constant(x_var), missing='drop')

This works - which as far as I understand the docs it enforcing HAC cov. However, I am not sure if I am calling it correctly

res = model.fit(cov_type='hac-panel', cov_kwds={'time': dat.index, 'maxlags': 11})
res.summary()

Here is where I have a problem. Let's say I want to also cluster by time, which I think should be something like this:

model.fit(cov_type='hac-panel', cov_kwds={'time': dat.index, 'groups': dat.index, 'maxlags': 11})

All help is really appreciated. Thank you very much in advance. Even pointing me to an example would be great - couldn't find anything.

I get this error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-2-74b3e662267b> in <module>
----> 1 model.fit(cov_type='hac-panel', cov_kwds={'time': dat.index, 'groups': dat.index, 'maxlags': 11})

~\anaconda3\lib\site-packages\statsmodels\regression\linear_model.py in fit(self, method, cov_type, cov_kwds, use_t, **kwargs)
    343                 self, beta,
    344                 normalized_cov_params=self.normalized_cov_params,
--> 345                 cov_type=cov_type, cov_kwds=cov_kwds, use_t=use_t)
    346         else:
    347             lfit = RegressionResults(

~\anaconda3\lib\site-packages\statsmodels\regression\linear_model.py in __init__(self, model, params, normalized_cov_params, scale, cov_type, cov_kwds, use_t, **kwargs)
   1555                 # TODO: warn or not?
   1556             self.get_robustcov_results(cov_type=cov_type, use_self=True,
-> 1557                                        use_t=use_t, **cov_kwds)
   1558         for key in kwargs:
   1559             setattr(self, key, kwargs[key])

~\anaconda3\lib\site-packages\statsmodels\regression\linear_model.py in get_robustcov_results(self, cov_type, use_t, **kwargs)
   2490             res.cov_params_default = sw.cov_nw_panel(self, maxlags, groupidx,
   2491                                                      weights_func=weights_func,
-> 2492                                                      use_correction=use_correction)
   2493             res.cov_kwds['description'] = descriptions['HAC-Panel']
   2494 

~\anaconda3\lib\site-packages\statsmodels\stats\sandwich_covariance.py in cov_nw_panel(results, nlags, groupidx, weights_func, use_correction)
    785     xu, hessian_inv = _get_sandwich_arrays(results)
    786 
--> 787     S_hac = S_nw_panel(xu, weights, groupidx)
    788     cov_hac = _HCCM2(hessian_inv, S_hac)
    789     if use_correction:

~\anaconda3\lib\site-packages\statsmodels\stats\sandwich_covariance.py in S_nw_panel(xw, weights, groupidx)
    723     S = weights[0] * np.dot(xw.T, xw)  #weights just for completeness
    724     for lag in range(1, nlags+1):
--> 725         xw0, xwlag = lagged_groups(xw, lag, groupidx)
    726         s = np.dot(xw0.T, xwlag)
    727         S += weights[lag] * (s + s.T)

~\anaconda3\lib\site-packages\statsmodels\stats\sandwich_covariance.py in lagged_groups(x, lag, groupidx)
    706 
    707     if out0 == []:
--> 708         raise ValueError('all groups are empty taking lags')
    709     #return out0, out_lagged
    710     return np.vstack(out0), np.vstack(out_lagged)

ValueError: all groups are empty taking lags

回答1:

was looking for an example and your was very helpful.

Only problem with your code seems to be using the same time index in

cov_kwds={'time': dat.index, 'groups': dat.index, 'maxlags': 11}

Basically, it accounts every unique unit from dat.index as a separate group, in your case every quarter. At the same time it uses that year as a time indicator so your group exists of all observations in the same quarter with time lenght one quarter. Since your time variable is just one period for each group there are no lags involved thus the error.

来源：https://stackoverflow.com/questions/61465483/python-statsmodels-robust-cov-type-hac-panel-issue

标签

python

statsmodels