Abstract :
In the mid 1980s attempts were made to devise a form of multivariate analysis which measured the tendency of species to occur towards one or other end of a transect in a way which was more than merely descriptive. While promising, the methods had conceptual and statistical difficulties, and were restricted to presence/absence information. However, they did have the potential to supplement existing multivariate methods by determining not only whether a community showed a significant skew in its distribution along a transect, but which species contributed most to that shift. This is an important issue where field work shows changes in species compositions along purported gradients, especially when it is claimed that the latter causes the former. Very often the linkage made between gradient and species changes is weak, with authors presenting assertions which may not have been properly established. This article outlines a way of strengthening this link. Skew values may be obtained quantitatively for each species; these measures are made relative to the whole community. A disadvantage of this is that bunching in the middle of a transect, or bunching anywhere relative to another physical parameter such as latitude, might be missed, so the following adds an analysis of ‘kite diagrams’ in which a measure of species clumping or dispersion is arrived at which is independent of other community components. Monte Carlo methods are used for determining significances of both skews and clumping. The method is applied to four published data sets, and is compared, in some cases, with results obtained by cluster analysis and multidimensional scaling.
