MLOC_SIM simulates an MLOC object (SOM with local models)
CALLS
[yest,pr] = mloc_sim(u,mloc,p,y0)
INPUTS
u: input data matrix (R x Q)
mloc: SOM structure with local models
p: data matrix with exogenous index (dynamic selectors) variables (Rp x Q)
y0: data matrix with the initial values of y
OUTPUTS
yest: matrix with estimated outputs
bmu: matrix with the sequence of mloc best matching units
Note:
The algorithm takes the following initial values (assuming ny>=nu)
[y0(k-1) y0(k-2) ... y0(k-ny)]
[ u(k-1) u(k-2) ... u(k-nu)]
and returns
[yest(1) yest(2) ... yest(k-ny), yest(k-ny+1) ... yest(Q)]
------------------------------- ------------------------
valores no predichos predicciones
That is, the "true predictions" are from element ny+1 of yest, so that:
yest(ny+1) is the one step prediction, for initial conditions y0(1:ny), u(1:ny)
yest(ny+2) is the two-step prediction, for initial conditions y0(1:ny), u(1:ny)
etc...
SEE ALSO
MLOC_LOC, MLOC_SOM, MLOC_PLANOS, MLOC_COMPARA