Question

I would like to compute fixed lag smoothing estimates of the state
variable in a state space model. These are estimations of the
state variable at one point in time given information for several
periods ahead, but not the whole sample.

Putting it in equations, in a State Space model usually the smoothed estimate of the state variable S, S t / T is computed. This is the estimate of the state at time t given the whole sample T. Using the Kalman filter, the filtered estimate S t /t can be computed as well.

I would like to compute S t / t+N, where N is a fixed number of
periods, and t+N < T.

Is anyone aware of an implementation of this fixed lag smoother in Kalman Filtering software?

Answer 1

I am not aware of any package in R computing fixed lag smoothing estimates. However, unless
you have to compute a great many of such estimates, I think that truncating the time series at t+N and computing the ordinary smoothed value would do what you want.