I am having difficulties understanding how to initialise an empty as well as as a known initial config model in statsmodels, here imported as
from statsmodels.tsa.regime_switching.markov_autoregression import MarkovAutoregression
# HMM fitting to determine the Parameters of the 2 state SWitching-AR(2,1) init_prob = np.array( [[0.8, 0.2], [0.2, 0.8]] ) MarkovAutoregression.initialize_known(probabilities=init_prob , tol=1e-08) SWARCH_model = MarkovAutoregression( bcox(ARCH['residuals']) , k_regimes=2, order=0, switching_ar=True, switching_variance=True, switching_exog=False) SWARCH = SWARCH_model.fit()
I get the following error:
MarkovAutoregression.initialize_known(probabilities=init_prob , tol=1e-08) TypeError: initialize_known() missing 1 required positional argument: 'self'
I have has similar experiences with the method:
MarkovAutoregression.initialize()
On the User Guide, it does not give any inputs to this method so I do not know the syntax needed to initialise an empty model.
Guide:
Advertisement
Answer
initialize_known is not a class method. You can call it like: SWARCH_model.initialize_known(…)