Constrained LLA to predict a chaotic time series

It should work with the noisy time series of limited length

4 files are included in the package:
(1) fs01.m - the 1st-order local linear approximation (LLA) very similar to algorithm by Farmer and Sidorowich (1987). Uses constrained linear least squares (lsqlin) and
(2) ts1.m - to reconstruct the vectors of the state space.
(3) Example.m - runs fs01 and ts1 to predict
(4) composite.txt - synthetic time series of decadal average sunspot number