Rethinking ecological inference: Density dependence in reef fishes

We use a meta-analysis of density dependence in reef fishes to evaluate how ecologists approach detection, inference, and estimation. We compared two groups of studies: those that detected effects of density on survival and those that did not. Distinctions between these groups have spawned heated debate about the processes that affect fish dynamics. Per capita effects of density were similar between the two groups, although total effects (and hence ambient density) were greater in studies that detected density effects. The majority of the variation in effects of density was not resolved by the classification of studies based on the authors' conclusions. These results suggest (1) that standard inferences based on null hypothesis tests may miss important sources of variation in effects and give rise to unnecessary debate; and (2) that estimation of effect sizes and model parameters (including their uncertainty) is a powerful alternative to detection of ecological processes.