Abstract

<jats:p>The robustness of site groupings produced by ordination (DECORANA) and classification (TWINSPAN) techniques to variations in the quality of the raw data was investigated, using two data sets on macroinvertebrate communities from the La Trobe River. Ordinations or classifications based on the presence or absence of species were not substantially different from those based on actual abundance levels. However, when taxonomic discrimination was reduced from the species (or genus) level to the family level, distortions occurred in the resulting ordinations and classifications. In addition, ordinations based on 10 replicates per sample were little different from those based on a subset of 5 or 6 of these replicates; fewer than 4 replicates did not adequately represent the patterns present in the full data set.</jats:p>