CLN: Use correct freq_to_periods

Use preferred one form tsatools

statsmodels/tsa/holtwinters.py

@@ -22,9 +22,8 @@
 from statsmodels.base.model import Results
 from statsmodels.base.wrapper import populate_wrapper, union_dicts, ResultsWrapper
 from statsmodels.tsa.base.tsa_model import TimeSeriesModel
+from statsmodels.tsa.tsatools import freq_to_period
 import statsmodels.tsa._exponential_smoothers as smoothers
-from statsmodels.tsa.x13 import _freq_to_period
-
 
 def _holt_init(x, xi, p, y, l, b):
     """Initialization for the Holt Models"""
@@ -502,10 +501,7 @@ def __init__(self, endog, trend=None, damped=False, seasonal=None,
         if self.seasoning:
             self.seasonal_periods = seasonal_periods
             if seasonal_periods is None:
-                try:
-                    self.seasonal_periods = _freq_to_period[self._index_freq.freqstr]
-                except (ValueError, TypeError, AttributeError):
-                    raise ValueError('Unable to detect seasonal periods.')
+                self.seasonal_periods = freq_to_period(self._index_freq)
             if self.seasonal_periods <= 1:
                 raise ValueError('seasonal_periods must be larger than 1.')
         else:

statsmodels/tsa/tests/test_holtwinters.py

@@ -286,11 +286,9 @@ def test_dampen_no_trend(seasonal):
 
 @pytest.mark.parametrize('seasonal', ('add', 'mul'))
 def test_invalid_seasonal(seasonal):
-    y = -np.ones(100)
+    y = pd.Series(-np.ones(100),index=pd.date_range('2000-1-1', periods=100, freq='MS'))
     with pytest.raises(ValueError):
         ExponentialSmoothing(y, seasonal=seasonal, seasonal_periods=1)
-    with pytest.raises(ValueError):
-        ExponentialSmoothing(y, seasonal=seasonal)
 
 
 def test_2d_data():