Stationary Models

Abstract

This chapter focuses on how to model stationary time series. It presents preliminary analysis techniques to investigate whether a process exhibits serial correlation. Stationary autocorrelated process data can often be modeled through an autoregressive moving average (ARMA) time series model. The identification of the particular model within this general class of models is determined by looking at the autocorrelation function (ACF) and the partial autocorrelation function (PACF). The chapter provides a brief description of PACF. Modeling the process and checking that the parameter estimates satisfy stationarity conditions is one way to test for stationarity. However, it would be desirable to have a simple exploratory tool to investigate whether a process is stationary without having to model it first. An important preliminary step in any data analysis is to consider the possibility of a (nonlinear) data transformation. Controlled Vocabulary Terms ARMA model; Autocorrelation; autocorrelation function; Data transformation; partial autocorrelation function; stationary process