Stochastic Identification Under Multiple Operating Conditions: Functionally Pooled VARMA Methods

Abstract

The identification of stochastic systems capable of operating under different conditions is addressed based on data records corresponding to each condition. The problem is important in various applications, and is tackled within a recently introduced, novel, Functional Pooling framework. The study focuses on the identification of Functionally Pooled Vector AutoRegressive Moving Average (FP–VARMA) models within this framework. Two–Stage Least Squares and Maximum Likelihood estimators are formulated, while model structure selection is postulated via a canonical correlation analysis scheme and information criteria. The performance characteristics of the identification approach are assessed via a Monte Carlo study.