On Some Ambiguities Associated with the Fitting of ARMA Models to Time Series

Written by:

Introduction

The ambiguities we discuss are of various kinds: In section 2, we illustrate some consequences for signal extraction of inappropriately using 'parsimony' to remove the ambiguity inherent in the selection of an ARMA model for an unobserved component of a stationary ARMA (p,p) process when only the spectrum of the observed process is known. In the remaining sections, we mainly look at consequences for various applications of only knowing a finite segment.

Page Last Revised - February 11, 2025