Design and analysis of experiments testing for biodiversity effects in ecology

It is now widely believed that biological diversity is good for the natural environment. One way that ecologists test this is to place random collections of species in mini-environments and then measure some outcome. Statisticians have been working with fresh-water ecologists to improve this in two ways. The first is that the subsets of species are carefully chosen, not random. The second is that a nested family of plausible models is fitted. The results of three experiments suggest that biodiversity can have no effect at all, but that there are other plausible underlying mechanisms.Implications for the design of such experiments, the understanding of the family of models, and the analysis of the data are discussed.