A direct approach to time-varying modelling

摘要: We consider the on-line identification of continuously adaptive autoregressive (AR) models for observed data records which are realizations non-stationary stochastic processes. Emphasis is placed on treatment arbitrary non-stationarities and use realistic assumptions in this operation. Because these two objectives, usual adaptation procedures or description techniques not well suited to problem, motivates investigation a new (direct) approach time-varying parameter estimation. The cost criterion considered constrained least squares functional incorporates with equal weight all instantaneous errors up current time observation. constraint specified from limited priori knowledge about nature non-stationarity, namely expected maximum rate change (MRC) model parameters.