ABSTRACT
A time series is a set of ordered observations, each of which has an associated observation time. Because of the ordering, observations are inherently not interchangeable unless the time series is effectively a random one, meaning that all the observations are independent values from the same distribution. It is therefore, in principle, only possible to use randomization to test a series for time structure against the null hypothesis that there is no structure at all. Procedures of this type, which are called tests for randomness or tests for independence, are reviewed by Gibbons (1986) and Madansky (1988, Chapter 3). Connor (1986) discusses applications associated with fossil records, and Chatfield (2003) provides a useful introductory text for time series analysis in general.
