Merge 98ea129 into 497819e

arch/tests/univariate/test_mean.py

@@ -14,6 +14,7 @@
 from pandas.util.testing import assert_frame_equal, assert_series_equal
 import pytest
 from scipy import stats
+from scipy.optimize import OptimizeResult
 import statsmodels.regression.linear_model as smlm
 import statsmodels.tools as smtools
 
@@ -439,7 +440,9 @@ def test_arch_arx(self):
         y = x.sum(1) + 3 * self.rng.randn(500)
 
         am = ARX(y=y, x=x)
-        am.fit(disp=DISPLAY).summary()
+        res = am.fit(disp=DISPLAY)
+        res.summary()
+        assert isinstance(res.optimization_result, OptimizeResult)
         am.volatility = ARCH(p=2)
         results = am.fit(update_freq=0, disp=DISPLAY)
         assert isinstance(results.pvalues, pd.Series)

arch/univariate/__init__.py

@@ -1,5 +1,6 @@
 from __future__ import absolute_import
 
+from arch.univariate import recursions_python
 from arch.univariate.distribution import (Distribution, GeneralizedError,
                                           Normal, SkewStudent, StudentsT)
 from arch.univariate.mean import (ARX, HARX, LS, ConstantMean, ZeroMean,
@@ -9,7 +10,6 @@
                                         FixedVariance, MIDASHyperbolic,
                                         RiskMetrics2006)
 
-from arch.univariate import recursions_python
 try:
     from arch.univariate import recursions
 except ImportError:

arch/univariate/base.py

@@ -1355,6 +1355,8 @@ class ARCHModelResult(ARCHModelFixedResult):
         Loglikelihood at estimated parameters
     is_pandas : bool
         Whether the original input was pandas
+    optim_output : OptimizeResult
+        Result of log-likelihood optimization
     fit_start : int
         Integer index of the first observation used to fit the model
     fit_stop : int
@@ -1607,6 +1609,18 @@ def convergence_flag(self):
         """
         return self._optim_output.status
 
+    @property
+    def optimization_result(self):
+        """
+        Information about the covergence of theloglikelihood optimization
+
+        Returns
+        -------
+        optim_result : OptimizeResult
+            Result from numerical optimization of the log-likelihood.
+        """
+        return self._optim_output
+
 
 def _align_forecast(f, align):
     if align == 'origin':

arch/utility/array.py

@@ -10,7 +10,8 @@
 import datetime as dt
 
 import numpy as np
-from pandas import DataFrame, NaT, Series, Timestamp, to_datetime, DatetimeIndex
+from pandas import (DataFrame, DatetimeIndex, NaT, Series, Timestamp,
+                    to_datetime)
 
 __all__ = ['ensure1d', 'parse_dataframe', 'DocStringInheritor',
            'date_to_index', 'cutoff_to_index', 'ensure2d', 'AbstractDocStringInheritor']

doc/source/changes/4.0.txt

@@ -1,5 +1,7 @@
 Changes since 4.8.1
 ===================
+- Added :func:`~arch.univariate.base.ARCHModelResult.optimization_result` to
+  simplify checking for convergence of the numerical optimizer (:issue:`292`).
 - Added `random_state` argument to :func:`~arch.univariate.HARX.forecast`
   to allow a :class:`~numpy.random.RandomState` object to be passed in when
   forecasting when `method='bootstrap'`. This allows the repeatable forecast to