Time-series Analysis

Abstract

A very important class of phenomena in science and engineering varies with time or some other parameter in an apparently unsystematic way. A few examples from engineering are radio noise, pressure variations in turbulence, heights of surface asperities, vibrations of vehicle structures, yearly flood levels. The fluctuations of these phenomena are essentially statistical in nature, and in the last generation or so a mathematical approach to their analysis has been worked out, mainly by communications engineers. Until recently the analysis was applied only to analogue signals which could be treated directly by electronic analogue techniques; this was because the computational work in treating a statistically representative series of discrete data was too great. This has all been changed by the advent of the digital computer, and time-series analysis is now an everyday tool of the practising scientist and engineer. The present chapter is an introduction to a few of the more elementary ideas of the subject, which we believe has not previously been treated at undergraduate level; even its simplest techniques are so powerful, however, and its use is becoming so widespread, that we do not feel a modern student-textbook on data analysis is complete without it.