I would like to run a panel regression (fixed-effect model) on a group of individuals, which are uniquely identified by province and city, across time t.
Code to create dataframe and run the regression
import numpy as np
import pandas as pd
from linearmodels import PanelOLS
data = {'y':[1,2,3,1,0,3],
'x1': [0,1,2,3,0,2],
'x2':[1,1,3,2,1,0],
't': ['2020-02-18', '2020-02-18', '2020-02-17', '2020-02-18', '2020-02-18', '2020-02-17'],
'province': ['A', 'A','A','B','B','B'],
'city': ['a','b','a','a','c','a']}
dataframe = pd.DataFrame (data, columns = ['y','x1', 'x2', 't', 'province', 'city'])
dataframe=dataframe.set_index(['t','province','city'], append=True)
mod = PanelOLS(dataframe.y, dataframe[['x1','x2']], entity_effects=True)
But I got an error which says "DataFrame input must have a MultiIndex with 2 levels."
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-33-eb5264bfefc9> in <module>
1 dataframe=dataframe.set_index(['t','province','city'], append=True)
----> 2 mod = PanelOLS(dataframe.y, dataframe[['x1','x2']], entity_effects=True)
C:\ProgramData\Anaconda3\lib\site-packages\linearmodels\panel\model.py in __init__(self, dependent, exog, weights, entity_effects, time_effects, other_effects, singletons, drop_absorbed)
1038 drop_absorbed: bool = False,
1039 ) -> None:
-> 1040 super(PanelOLS, self).__init__(dependent, exog, weights=weights)
1041
1042 self._entity_effects = entity_effects
C:\ProgramData\Anaconda3\lib\site-packages\linearmodels\panel\model.py in __init__(self, dependent, exog, weights)
224 weights: Optional[PanelDataLike] = None,
225 ) -> None:
--> 226 self.dependent = PanelData(dependent, "Dep")
227 self.exog = PanelData(exog, "Exog")
228 self._original_shape = self.dependent.shape
C:\ProgramData\Anaconda3\lib\site-packages\linearmodels\panel\data.py in __init__(self, x, var_name, convert_dummies, drop_first, copy)
198 if len(x.index.levels) != 2:
199 raise ValueError(
--> 200 "DataFrame input must have a " "MultiIndex with 2 levels"
201 )
202 if isinstance(self._original, (DataFrame, PanelData, Series)):
ValueError: DataFrame input must have a MultiIndex with 2 levels
As a solution, instead of doing
dataframe=dataframe.set_index(['t','province','city'], append=True)
I do this
dataframe=dataframe.set_index(['t'], append=True)
This will allow the model to go through. But I do not know why. In this case, I am using two columns to identify the group. What if I need three columns to identify my groups? How does python differentiate between the ID and x variables?
According to the author of linearmodels, I need to have a single entity,
import numpy as np
import pandas as pd
from linearmodels import PanelOLS
data = {'y':[1,2,3,1,0,3],
'x1': [0,1,2,3,0,2],
'x2':[1,1,3,2,1,0],
't': pd.to_datetime(['2020-02-18', '2020-02-18', '2020-02-17', '2020-02-18', '2020-02-18', '2020-02-17']),
'province': ['A', 'A','A','B','B','B'],
'city': ['a','b','a','a','c','a']}
dataframe = pd.DataFrame (data, columns = ['y','x1', 'x2', 't', 'province', 'city'])
dataframe["city-provence"] = [(c,p) for c,p in zip(dataframe.city, dataframe.province)]
dataframe = dataframe.set_index(["city-provence","t"])
y x1 x2 province city
city-provence t
(a, A) 2020-02-18 1 0 1 A a
(b, A) 2020-02-18 2 1 1 A b
(a, A) 2020-02-17 3 2 3 A a
(a, B) 2020-02-18 1 3 2 B a
(c, B) 2020-02-18 0 0 1 B c
(a, B) 2020-02-17 3 2 0 B a