Missing Data

All of the models can handle missing data. For performance reasons, the default is not to do any checking for missing data. If, however, you would like for missing data to be handled internally, you can do so by using the missing keyword argument. The default is to do nothing

In [1]: import statsmodels.api as sm

In [2]: data = sm.datasets.longley.load()

In [3]: data.exog = sm.add_constant(data.exog)

# add in some missing data
In [4]: missing_idx = np.array([False] * len(data.endog))

In [5]: missing_idx[[4, 10, 15]] = True

In [6]: data.endog[missing_idx] = np.nan

In [7]: ols_model = sm.OLS(data.endog, data.exog)

In [8]: ols_fit = ols_model.fit()

In [9]: print(ols_fit.params)
const     NaN
GNPDEFL   NaN
GNP       NaN
UNEMP     NaN
ARMED     NaN
POP       NaN
YEAR      NaN
dtype: float64

This silently fails and all of the model parameters are NaN, which is probably not what you expected. If you are not sure whether or not you have missing data you can use missing = ‘raise’. This will raise a MissingDataError during model instantiation if missing data is present so that you know something was wrong in your input data.

In [10]: ols_model = sm.OLS(data.endog, data.exog, missing='raise')
---------------------------------------------------------------------------
MissingDataError                          Traceback (most recent call last)
Cell In [10], line 1
----> 1 ols_model = sm.OLS(data.endog, data.exog, missing='raise')

File /usr/lib/python3/dist-packages/statsmodels/regression/linear_model.py:906, in OLS.__init__(self, endog, exog, missing, hasconst, **kwargs)
    903     msg = ("Weights are not supported in OLS and will be ignored"
    904            "An exception will be raised in the next version.")
    905     warnings.warn(msg, ValueWarning)
--> 906 super(OLS, self).__init__(endog, exog, missing=missing,
    907                           hasconst=hasconst, **kwargs)
    908 if "weights" in self._init_keys:
    909     self._init_keys.remove("weights")

File /usr/lib/python3/dist-packages/statsmodels/regression/linear_model.py:733, in WLS.__init__(self, endog, exog, weights, missing, hasconst, **kwargs)
    731 else:
    732     weights = weights.squeeze()
--> 733 super(WLS, self).__init__(endog, exog, missing=missing,
    734                           weights=weights, hasconst=hasconst, **kwargs)
    735 nobs = self.exog.shape[0]
    736 weights = self.weights

File /usr/lib/python3/dist-packages/statsmodels/regression/linear_model.py:190, in RegressionModel.__init__(self, endog, exog, **kwargs)
    189 def __init__(self, endog, exog, **kwargs):
--> 190     super(RegressionModel, self).__init__(endog, exog, **kwargs)
    191     self._data_attr.extend(['pinv_wexog', 'wendog', 'wexog', 'weights'])

File /usr/lib/python3/dist-packages/statsmodels/base/model.py:267, in LikelihoodModel.__init__(self, endog, exog, **kwargs)
    266 def __init__(self, endog, exog=None, **kwargs):
--> 267     super().__init__(endog, exog, **kwargs)
    268     self.initialize()

File /usr/lib/python3/dist-packages/statsmodels/base/model.py:92, in Model.__init__(self, endog, exog, **kwargs)
     90 missing = kwargs.pop('missing', 'none')
     91 hasconst = kwargs.pop('hasconst', None)
---> 92 self.data = self._handle_data(endog, exog, missing, hasconst,
     93                               **kwargs)
     94 self.k_constant = self.data.k_constant
     95 self.exog = self.data.exog

File /usr/lib/python3/dist-packages/statsmodels/base/model.py:132, in Model._handle_data(self, endog, exog, missing, hasconst, **kwargs)
    131 def _handle_data(self, endog, exog, missing, hasconst, **kwargs):
--> 132     data = handle_data(endog, exog, missing, hasconst, **kwargs)
    133     # kwargs arrays could have changed, easier to just attach here
    134     for key in kwargs:

File /usr/lib/python3/dist-packages/statsmodels/base/data.py:700, in handle_data(endog, exog, missing, hasconst, **kwargs)
    697     exog = np.asarray(exog)
    699 klass = handle_data_class_factory(endog, exog)
--> 700 return klass(endog, exog=exog, missing=missing, hasconst=hasconst,
    701              **kwargs)

File /usr/lib/python3/dist-packages/statsmodels/base/data.py:72, in ModelData.__init__(self, endog, exog, missing, hasconst, **kwargs)
     70     self.formula = kwargs.pop('formula')
     71 if missing != 'none':
---> 72     arrays, nan_idx = self.handle_missing(endog, exog, missing,
     73                                           **kwargs)
     74     self.missing_row_idx = nan_idx
     75     self.__dict__.update(arrays)  # attach all the data arrays

File /usr/lib/python3/dist-packages/statsmodels/base/data.py:286, in ModelData.handle_missing(cls, endog, exog, missing, **kwargs)
    283     return combined, []
    285 elif missing == 'raise':
--> 286     raise MissingDataError("NaNs were encountered in the data")
    288 elif missing == 'drop':
    289     nan_mask = ~nan_mask

MissingDataError: NaNs were encountered in the data

If you want statsmodels to handle the missing data by dropping the observations, use missing = ‘drop’.

In [11]: ols_model = sm.OLS(data.endog, data.exog, missing='drop')

We are considering adding a configuration framework so that you can set the option with a global setting.