Tests of hypotheses implying a longitudinal association between or among time series have become increasingly common in the psychological literature. These tests, however, vary greatly in analytic approach and therefore focus attention on methodological rather than on substantive issues. The authors believe that the lack of methodological convergence is due to the fact that time-series analyses have historically evolved to meet forecasting rather than hypothesis testing needs. This article offers a scheme by which researchers can choose among time-series methods to test various classes of hypotheses. The scheme is derived from experimental logic and should help psychologists converge on time-series conventions that will allow attention to focus on substance rather than method. Empirical applications of the scheme are illustrated with several examples of psychiatric admissions series.

No abstract

No abstract

No abstract