A Web-based software for randomization tests of cluster analysis of invertebrate biodiversity in a rice ecosystem

As an important analytical tool, cluster analysis is widely used in ecological research–e.g., community classification, biological evolution analyses, and biogeographic comparisons (Krebs 1989). Dozens of algorithms for cluster analysis have been developed for common use or special purposes (Zhang and Fang 1982). Most of them, however, do not use appropriate statistical tests in the computation procedures. Thus, we are not able to evaluate statistically the confidence of the classifications in the cluster analysis. Classical statistics can be used to answer the above question when statistical assumptions on data have been met. For example, are the individuals randomly sampled from the population of interest? Do clustered individuals come from different populations or groups that share equal population standard deviations or means? Do the values coincide with a normal distribution or with other known distributions (Manly 1997)?