Linear Models and Time-Series Analysis

9ARMA Model Identification

There are two things you are better off not watching in the making: sausages and econometric estimates.

(Edward Leamer, 1983, p. 37)

Establishing plausible values of and associated with an ARMA() model corresponding to a given set of time‐series data constitutes an important part of what is referred to as (univariate time series) model identification, a term and procedure popularized by the highly influential book on time‐series analysis by the prolific George Box and Gwilym Jenkins, the first of which appeared in 1970; see Box et al. (2008 ).¹ Other aspects of the Box and Jenkins paradigm include parameter estimation and out‐of‐sample forecasting, which were covered in previous chapters.

9.1 Introduction

One reason why Akaike does not accept the problem of ARMA order selection as that of estimating an unknown true order, (, ), say, is that there is no fundamental reason why a time series need necessarily follow a ‘true’ ARMA model.

(Raj J. Bhansali, 1993, ...