[A,D,CC,F,U,E] = filter(A,D,Range,...)
A [ FAVAR ] - Estimated FAVAR object.
D [ struct | tseries ] - Input database or tseries object with the FAVAR observables.
Range [ numeric ] - Filter date range.
A [ FAVAR ] - FAVAR object.
D [ struct ] - Output database or tseries object with the FAVAR observables.
CC [ struct | tseries ] - Re-estimated common components in the observables.
F [ tseries ] - Re-estimated common factors.
U [ tseries ] - Re-estimated idiosyncratic residuals.
E [ tseries ] - Re-estimated structural residuals.
'cross=' [ true | false | numeric ] - Run the filter with the off-diagonal elements in the covariance matrix of idiosyncratic residuals; if false all cross-covariances are reset to zero; if a number between zero and one, all cross-covariances are multiplied by that number.
'invFunc=' [ 'auto' | function_handle ] - Inversion method for the FMSE matrices.
'meanOnly=' [ true | false ] - Return only mean data, i.e. point estimates.
'persist=' [ true | false ] - If filter or forecast is used with 'persist=' set to true for the first time, the forecast MSE matrices and their inverses will be stored; subsequent calls of the filter or forecast functions will re-use these matrices until filter or forecast is called.
'output=' [ 'auto' | 'dbase' | 'tseries' ] - Format of output data.
'tolerance=' [ numeric | 0 ] - Numerical tolerance under which two FMSE matrices computed in two consecutive periods will be treated as equal and their inversions will be re-used, not re-computed.
It is the user's responsibility to make sure that filter and forecast called with 'persist=' set to true are valid, i.e. that the previously computed FMSE matrices can be really re-used in the current run.