Recursive multivariate time-series analysis with the inputs split into subsets with separate matrix transfer functions
P.F. Davis · International Journal of Control · 1982
This report describes a development in the recursive analysts of multivariate time-series using instrumental variables to relate inputs and outputs. At each sampling instant the set of inputs is split into discrete subsets, each having a separate matrix transfer function relationship with an associated vector of instrumental variables. The analysis is simplified when the input subsets are mutually uncorrelated. The vector of outputs is related to the sum of the concurrent vectors of instrumental variables. Tests of the new algorithm on simulated time-series resulted in satisfactory parameter estimates in most cases.