Research work · 2007

Autoregressive Integrated Moving Average ( ARIMA ) Modeling

Alberto Luceño, Daniel Peña

Encyclopedia of Statistics in Quality and Reliability

Abstract

Abstract ARIMA (autoregressive integrated moving average) modeling is a method to approach the dynamic behavior of a process that generates a time series of successive valuesY′ = (y1, …,yn) of a magnitudeYobserved at equispaced periods {S, …,nS}, whereSis a sampling interval. ARIMA is composed of three parts, namely, AR, I, and MA, that can stand alone or be combined at will. Hence there are white noise models (without any part) as well as AR, I (e.g., random walk), MA, ARI, ARMA, IMA, and ARIMA models. When the I part is included, the process is nonstationary. Differentiation of a nonstationary ARIMA process leads to a stationary ARMA process. ARIMA models belong to the class of linear time series models and often assume that vectorY′ is jointly normally distributed, possibly after transformation. ARIMA modeling is usually implemented by identifying the order of the AR, I, and MA parts, estimating the AR and MA parameters, and diagnosing the adequacy of the hypotheses. Short‐term forecasting with ARIMA models is often difficult to beat in the quality context by using nonlinear and/or nonnormal models.

DOI: 10.1002/9780470061572.eqr276