00430 )f097db8548f2de1e3b627fe4f9c1b6

20

A Graphical Aid for Analyzing Autocorrelated Dynamical Systems

Edmund L. Russell III AcfaancedMicro Devices Austin, Texas

Abstract

Autocorrelated dynamical systems of short duration or with rapidly changing underlying models present a major modeling challenge to statisticians. In this paper, an interactive graphical exploratory technique is presented that can assist the statistician with: determining the order of the underlying model, determining if the underlying model shifted, and identifying outliers. Synthetic data, actual in-process data from industry, and lastly an example using the methodology to assist in fitting a model will be examined.

Introduction

Over the past few years researchers in a variety of disciplines have become highly interested in dynamical systems. Dynamical systems are recursive systems that map back into their own domain. The relationship

435