use_boxcox parameter in holt winters

[callmegar]

with this grid:

hwes = {
'trend':['add','mul'],
'seasonal':['add','mul'],
'damped_trend':[True,False],
'initialization_method':[None,'estimated','heuristic'],
'use_boxcox':[True,False],
'seasonal_periods':[168],
}

I get this result which I do not understand as it is True or False:

`File ~/.local/share/virtualenvs/scalecast-0HQw5DtN/lib/python3.10/site-packages/statsmodels/tsa/holtwinters/model.py:337, in ExponentialSmoothing._boxcox(self)
335 y = boxcox(self._y, self._use_boxcox)
336 else:
--> 337 raise TypeError("use_boxcox must be True, False or a float.")
338 return y

TypeError: use_boxcox must be True, False or a float.`

I think the issue may be here:
File ~/.local/share/virtualenvs/scalecast-0HQw5DtN/lib/python3.10/site-packages/statsmodels/tsa/holtwinters/model.py:291, in ExponentialSmoothing.__init__(self, endog, trend, damped_trend, seasonal, seasonal_periods, initialization_method, initial_level, initial_trend, initial_seasonal, use_boxcox, bounds, dates, freq, missing) 289 self._use_boxcox = use_boxcox 290 self._lambda = np.nan --> 291 self._y = self._boxcox() 292 self._initialize() 293 self._fixed_parameters = {}

[mikekeith52]

This isn't the first time I've seen this issue but I really started digging into it after seeing this issue. Turns out, True or False gets converted from a bool to a numpy.bool_ type between the time the model is tuned and tested. This causes the check kwargs['use_boxcox'] is True to return False, thus raising the issue which originates from statsmodels. Incredible. I will implement a fix in 0.15.3.

[mikekeith52]

Please try it out and let me know.