Comparing categorical and continuous ecological analyses: Effects of "wave exposure" on rocky shores

Abstract

Development of general theories and subsequent empirical testing are fundamental ingredients in ecological science. The progress of such efforts is determined by the logical coherence among central concepts, theories, and predictions on one hand, and experimental design, statistical analyses, and interpretation of results on the other. Here, we specifically explore an example of how differences in the way ecological concepts are defined lead to differences in the formulation and statistical testing of hypotheses and ultimately to differences in conclusions about the relative importance of ecological processes. In marine intertidal habitats, the notion that wave exposure has an important structuring role is widely agreed upon. Nevertheless, generalizations about its effects and use for accurate prediction of assemblages are often limited. This may partly be explained by the frequent use of categorical rather than quantitative definitions of wave exposure. We compared the conclusions about the importance of wave exposure from (1) analyses of variance based on relative classification of wave exposure and geographic location to those of (2) regression analyses based on continuous measures from 16 locations on the Swedish west coast. Variability in richness was substantially better explained by the regression analyses, while for the cover of individual taxa there was no consistent difference between the two analytical approaches in terms of explained variability. The two approaches detected significant spatial patterns for the same taxa, but conclusions about the nature of these patterns were often divergent. Categorical analyses of relative measures of wave exposure and geographic location indicated that interactive effects and differences between geographic areas were predominant. Regression analyses of absolute, continuous measures suggested that mean significant wave height was a better predictor than geographic location and interactive terms. Thus, the choice of definition of wave exposure has important consequences for how causes of spatial patterns of intertidal assemblages are perceived. Categorical analyses appear to provide clearer indications as to which factors are important while the use of continuous predictors sometimes provides a better fit to the data. The consequences of these findings are discussed in the context of rocky shore ecology as well as in a general perspective of ecological models, hypotheses, experimentation, and analysis. © 2005 by the Ecological Society of America.