TIME-SERIES ANALYSIS

The extraction of information from a series of measurements of a phenomenon made at intervals through time. As time-series data have a natural temporal ordering of observations, their analysis is distinct from other common data-analysis methods, which often have no natural ordering. Timeseries data sets are often referred to as either deterministic or stochastic signals, with each of these allied with two main types of methods of analysis. Frequencydomain methods are suitable for deterministic signals, where the analyses focus on the frequencies occurring in the data sets. Time-domain methods, which focus on how a signal changes over time, are most suited to stochastic signals. Examples of the former include spectral analysis and wavelet analysis. The latter include autocorrelation and cross-correlation analysis. Often time-series analysis is conducted for forecasting and prediction based on models, of which there are three broad classes: autoregressive models, integrated models and moving-average models. In the field of environmental change, time-series analysis is widely applied in, for example, dendrochronology, phenology, palaeohydrology, palae oclimatology and studies of climatic change.

[See alsoautoregressive (ar) modelling, filtering, spectral analysis, stochasticity]