4.8.6.4.2.1.7. statsmodels.tsa.kalmanf.kalmanfilter.StateSpaceModel.fit_kalman

StateSpaceModel.fit_kalman(start_params, xi10, ntrain=1, F=None, A=None, H=None, Q=None, R=None, method='bfgs', penalty=True, upperbounds=None, lowerbounds=None)

Parameters:

method : str

Only “bfgs” is currently accepted.

start_params : array-like

The first guess on all parameters to be estimated. This can be in any order as long as the F,A,H,Q, and R functions handle the parameters appropriately.

xi10 : arry-like

The (r x 1) vector of initial states. See notes.

F,A,H,Q,R : functions or array-like, optional

If functions, they should take start_params (or the current value of params during iteration and return the F,A,H,Q,R matrices). See notes. If they are constant then can be given as array-like objects. If not included in the state-space representation then can be left as None. See example in class docstring.

penalty : bool,

Whether or not to include a penalty for solutions that violate the bounds given by lowerbounds and upperbounds.

lowerbounds : array-like

Lower bounds on the parameter solutions. Expected to be in the same order as start_params.

upperbounds : array-like

Upper bounds on the parameter solutions. Expected to be in the same order as start_params