Merge pull request #190 from bashtage/summary-repr

ENH: Switch repr of results to summary

Author: bashtage

arch/tests/univariate/test_mean.py

@@ -395,7 +395,12 @@ def test_arch_arx(self):
                       'omega', 'alpha[1]', 'alpha[2]'])
 
         am = ARX(y=y, lags=2, x=x)
-        am.fit(disp=DISPLAY).summary()
+        res = am.fit(disp=DISPLAY)
+        summ = res.summary().as_text()
+        repr = res.__repr__()
+        assert str(hex(id(res))) in repr
+        assert summ[:10] == repr[:10]
+
         am.volatility = ARCH(p=2)
         results = am.fit(update_freq=0, disp='off')
         assert isinstance(results.pvalues, pd.Series)

arch/univariate/base.py

@@ -722,7 +722,23 @@ def forecast(self, params, horizon=1, start=None, align='origin', method='analyt
         raise NotImplementedError('Subclasses must implement')
 
 
-class ARCHModelFixedResult(object):
+class _SummaryRepr(object):
+    """Base class for returning summary as repr and str"""
+
+    def summary(self):
+        return Summary()
+
+    def __repr__(self):
+        out = self.__str__() + '\n'
+        out += self.__class__.__name__
+        out += ', id: {0}'.format(hex(id(self)))
+        return out
+
+    def __str__(self):
+        return self.summary().as_text()
+
+
+class ARCHModelFixedResult(_SummaryRepr):
     """
     Results for fixed parameters for an ARCHModel model
 
@@ -1596,11 +1612,11 @@ class ARCHModelForecast(object):
     simulations : ARCHModelForecastSimulation
         Object containing detailed simulation results if using a simulation-based method
     """
+
     def __init__(self, index, mean, variance, residual_variance,
                  simulated_paths=None, simulated_variances=None,
                  simulated_residual_variances=None, simulated_residuals=None,
                  align='origin'):
-
         mean = _format_forecasts(mean, index)
         variance = _format_forecasts(variance, index)
         residual_variance = _format_forecasts(residual_variance, index)

arch/univariate/recursions.pyx

@@ -14,6 +14,8 @@ cdef extern from 'math.h':
 cdef extern from 'float.h':
     double DBL_MAX
 
+cdef double LNSIGMA_MAX = log(DBL_MAX)
+
 @cython.boundscheck(False)
 @cython.wraparound(False)
 @cython.cdivision(True)
@@ -265,14 +267,13 @@ def egarch_recursion(double[:] parameters,
             else:
                 lnsigma2[t] += parameters[loc] * lnsigma2[t - 1 - j]
             loc += 1
+        if lnsigma2[t] > LNSIGMA_MAX:
+            lnsigma2[t] = LNSIGMA_MAX
         sigma2[t] = exp(lnsigma2[t])
         if sigma2[t] < var_bounds[t, 0]:
             sigma2[t] = var_bounds[t, 0]
         elif sigma2[t] > var_bounds[t, 1]:
-            if sigma2[t] > DBL_MAX:
-                sigma2[t] = var_bounds[t, 1] + 1000
-            else:
-                sigma2[t] = var_bounds[t, 1] + log(sigma2[t] / var_bounds[t, 1])
+            sigma2[t] = var_bounds[t, 1] + log(sigma2[t]) - log(var_bounds[t, 1])
         std_resids[t] = resids[t] / sqrt(sigma2[t])
         abs_std_resids[t] = fabs(std_resids[t])
 

arch/univariate/recursions_python.py

@@ -13,6 +13,8 @@
 __all__ = ['harch_recursion', 'arch_recursion', 'garch_recursion',
            'egarch_recursion']
 
+LNSIGMA_MAX = np.log(np.finfo(np.double).max) - .1
+
 
 def harch_recursion_python(parameters, resids, sigma2, lags, nobs, backcast,
                            var_bounds):
@@ -225,14 +227,13 @@ def egarch_recursion_python(parameters, resids, sigma2, p, o, q, nobs,
             else:
                 lnsigma2[t] += parameters[loc] * lnsigma2[t - 1 - j]
             loc += 1
+        if lnsigma2[t] > LNSIGMA_MAX:
+            lnsigma2[t] = LNSIGMA_MAX
         sigma2[t] = np.exp(lnsigma2[t])
         if sigma2[t] < var_bounds[t, 0]:
             sigma2[t] = var_bounds[t, 0]
         elif sigma2[t] > var_bounds[t, 1]:
-            if not np.isinf(sigma2[t]):
-                sigma2[t] = var_bounds[t, 1] + log(sigma2[t] / var_bounds[t, 1])
-            else:
-                sigma2[t] = var_bounds[t, 1] + 1000
+            sigma2[t] = var_bounds[t, 1] + log(sigma2[t]) - log(var_bounds[t, 1])
         std_resids[t] = resids[t] / np.sqrt(sigma2[t])
         abs_std_resids[t] = np.abs(std_resids[t])