Contributions to the Study of Oscillatory Time-Series

Abstract

THE classical approach to time-series was a search for hidden periodicities, observational error being considered the only obstacle to accurate prediction, and there is no doubt that it was the striking success of the associated methods in astronomy and tidal theory which led to the confident belief that they might be equally relevant to the study of social phenomena. Now, however, in this and many other fields one prefers, to work with a quite different model, in which the time-series is generated by a linear operator supplied with a random input. (A pleasing example is the music made by a sea shell.) The resulting series, though in general oscillatory, will not be periodic, and the possibility of prediction will be severely limited by the continued occurrence of random changes in phase and amplitude. Contributions to the Study of Oscillatory Time-Series By Maurice G. Kendall. (National Institute of Economic and Social Research Occasional Papers, No. 9.) Pp. viii + 76. (Cambridge: At the University Press, 1946.) 7s. 6d. net.