Intervention analysis with nonlinear dependent noise variation

Investigations of nonlinear fittings in combination with interventions have not been found in the literature so far. Because of the complexities of the procedures and intractability of the model identification and estimation of parameters of nonlinear models, the study has not developed. Recently the authors of this work developed a technique to fit a time series using a nonlinear model called the Quadratic Volterra Type (QVT) model (see Sarkar and Kartikeyan, 1987). Its methodology is quite tractable and easily amenable to the present context where nonlinearity is a dominant factor in studying the impact of interventions. We present methods of studying such nonlinear time series with three different kinds of intervention. Examples with naturally occurring series and with simulated data are presented to illustrate our techniques.