Time Series Analysis: Forecasting and Control, Fourth Edition

CHAPTER THIRTEEN

Intervention Analysis Models and Outlier Detection

Time series are often affected by special events or circumstances such as policy changes, strikes, advertising promotions, environmental regulations, and similar events, which we shall refer to as intervention events. In this chapter we describe the method of intervention analysis, which can account for the expected effects of these interventions. For this, the transfer function models of the previous chapters are used, but in the intervention analysis model the input series will be in the form of a simple pulse or step indicator function to indicate the presence or absence of the event. Initially, it will be assumed that the timing of the intervention event is known. Later, methods for the related problem of detection of outlying or unusual behavior in a time series at an unknown point of time are also discussed.

13.1 INTERVENTION ANALYSIS METHODS

13.1.1 Models for Intervention Analysis

In the setting of intervention analysis, it is assumed that an intervention event has occurred at a known point in time T of a time series. It is of interest to determine whether there is any evidence of a change or effect, of an expected kind, on the time series Y_t under study associated with the event. We consider the use of transfer function models to model the nature of and estimate the magnitude of the effects of the intervention, and hence to account for the possible unusual behavior in the time series related to the event. ...