I am using mgarch package in python, and I am fitting a MGARCH in volatility like this:
returns = get_returns() # dataframe with timestamp as index, tickers as columns
train, test = train_test_split(returns, train_size=len(returns)*3//4)
dist = 't'
vol = mgarch.mgarch(dist)
vol.fit(train)
predicted_vol = np.sqrt(np.diag(vol.predict(10)['cov']))
As expected, the volatility converges to the mean. But this only allows to predict n steps ahead with the same last observation of training set.
But I'd like to predict 10d ahead with new data point every step. How can I do please?
