ABSTRACT
A variety of analytical techniques have been applied to biological and physiological data suspected of having inherent periodicity. Choice among methods, and proponency for one or another technique, depends on the relative values given to the three often competitive goals of such analyses: (1) inferential validity; (2) descriptive utility; and (3) ease of data reduction and communication of results. The primary and necessary goal of time-series analysis is to demonstrate the presence of rhythmicity (as opposed to its absence), for which any of several methods suffice. The null hypothesis is straightforward: Observed variability of a data ensemble is randomly distributed with respect to time. Rejection of this hypothesis, with derived statistical estimates of probability, may be achieved by application of least squares curve fitting (Halberg, Tong, & Johnson, 1967); spectral analysis (Dumermuth & Fluehler, 1967; Gevins, Yaeger, Diamond, Spire, Zeitlin, & Gevins, 1975; Gotman, Skuce, Thompson, Gloor, Ives, & Ray, 1973; Walter, 1963, 1968a); correllography (Walter, 1963); periodography (Bliss, 1970; Orr & Hoffman, 1974; Walter, 1963). These and variant methods likewise may be used to provide essential parameters (frequency, phase, and amplitude) of rhythmic content by which natural or experimental effects are assessed. In most instances statistics and displays of analytical results have become standardized and readily understandable.
