Time-series analysis. Use of autocorrelation as an analytic strategy for describing pattern and change.

Although graphic representation of time-series data is one method to describe patterns of phenomena, autocorrelations, correlograms, and plots of the autocorrelation function provide descriptive and statistical methods that reveal the structure of a deterministic cycle component within the time-series. The autocorrelation function provides a method for the investigator to test hypotheses about the nature and pattern of relationships between measured and latent variables. Patterns of phenomena can be analyzed statistically using objectively provided criteria. The autocorrelation function can also be used to understand the change in response or behavioral patterns following an experimental intervention. Traditional group comparison designs, using cross-sectional data collection strategies cannot identify differences within the individual nor can they identify patterns of behavior or structural patterns within the data. Autocorrelation and cross-correlation become threats to statistical validity when conventional methods are used; however, in time-series analysis, autocorrelation allows close scrutiny of the pattern of response within an individual across a time-dimension.