The statistical analysis of tidal rhythms: Tests of the relative effectiveness of five methods using model simulations and actual data

Endogenous rhythms in intertidal organisms are often very complex and imprecise. Thus, confusing results are sometimes obtained when applying various interpretive analytical techniques. In an attempt to resolve this problem, ten different models representing typical organismic tidal‐rhythm displays were created and examined with five different inferential statistical techniques. The exercise was designed to test the relative effectiveness of these techniques in detecting the presence of known cycles in the models, and estimating their period lengths. The same comparison was then repeated on sets of animal‐derived data.

All of the five methods had their merits, but, depending on the model being examined, the results from the various methods were not identical. Three of the techniques produce harmonics, making data that contain multiple periods especially difficult to decipher. Often both tidal and circadian periods are displayed by shore dwellers; all five methods were able to find these two periods. But when the difference in circa period length was close, only one technique (array analysis) could make a distinction. This technique was also the only one able to handle data in which the period was not constant. Interestingly, this simplest of methods is probably the best all‐round method of discovery. Many more subtle, but important, differences were also noted, and it is recommended that more than one method always be used to ensure accuracy.