ABSTRACT
Time series analysis as an area of applied statistics arose in part because of a need for accurate prediction at varying distances into the future; indeed much work in time series in higher dimensions, such as mappings onto the surface of a sphere, arise in weather forecasting. The mathematics are specialized, difficult for a beginner, and have no immediately obvious psychological application (Mendel & Gieseking, 1971). But describing the autocorrelated structure in detail from a series of data points, univariate or multivariate, is not an end in itself even though it provides insights about how a system functions that are not available from direct inspection nor from static analysis. Reducing a messy record to a partitioning of stationary components, linearly or nonlinearly added, and a residual independent noise, is a means of “getting a handle” on process structure in order to increase our confidence in statements about what will happen next, or some time ahead.
