Blackwell Reference Online

Extract

Time series analysis can be used in two general situations: (1) forecasting and (2) exploring the nature of some event represented by a set of points or observations, with both techniques serving the purpose for establishing theory that can, at some point, represent future events. Time series analyses are often used in business settings by forecasting stock, commodity, and product valuing. Time series is less often used in educational research; therefore, it is this perspective that will be used in all examples to contextualize possible research scenarios where time series analyses would be appropriate. Among the myriad techniques subsumed by the term time series analyses are autocorrelation, trend, and seasonal variation which all help in the quest to understand the underlying structure or the fit of a theoretical model. Just as with the general linear model, time series can handle single or multiple dependent variables. Some techniques for fitting a time series include Box-Jenkins univariate and multivariate, and Holt-Winters. The fit techniques, similar to the way classical measurement attempts to differentiate between true and unsystematic error score for each item, attempt to differentiate between data points that are and are not useful in helping to predict future events. Therefore, time series analyses incorporate procedures for dealing with these erroneous data points. Specifically, ... log in or subscribe to read full text