MEPS 214:307-310 (2001)  -  doi:10.3354/meps214307

ABSTRACT: Lunar cycles are commonly observed in the movement, feeding and reproduction of marine fishes and invertebrates. The statistical techniques employed to examine these cycles are unstandardized, complex, and typically lacking in statistical power. Here we suggest a simple, sensitive and robust alternative for the detection of cyclical patterns: periodic regression. We use Monte Carlo simulation to demonstrate that periodic regression is more powerful and less sensitive to missing data than categorical ANOVA (the most commonly employed technique in the literature). Finally, we use real seahorse bycatch data to show that periodic regression is superior to categorical ANOVA for the detection and description of more complex cycles. We encourage researchers to use periodic regression in the analysis of lunar cycles and other cyclical patterns of known period.

Lunar cycles · Statistical techniques · Periodic regression