A simple graphical method, COREM, is described for the comparison of quantitative resemblance measures. The basis of the method is the ordered comparison case series (OCCAS) composed of relev6 pairs ranked in decreasing order of similarity. For a given OCCAS, values of given resemblance functions are calculated and plotted in COREM.
Delta resemblance (DR) is defined as the rate of change between two neighbouring relev6 pairs in the OCCAS. High mean delta resemblance (DR) of a resemblance measure indicates a high power of resolution and small standard deviation (SDDR) is indicative of linearity. A measure of efficiency combines DR and SDDR in the coefficient of variation for delta resemblance (CVDR). An efficient measure shows balanced sensitivity to the common and differential species. With the method COREM locally optimal resemblance measures may be selected for any given set of relev6s which produce the sharpest separation of clusters. A reasonable compromise is to use types of measures, such as the similarity ratio (SIM6), Czekanowski coefficient (SIM8), or floristic similarity (SIM9), which have the best chance of being successful in most relev6 sets. A study of transformation functions on Braun-Blanquet values with the help of COREM demonstrates that the combined transformation is the most advantageous for clustering. It is shown that standardization by relev6s, or conversion to unit sum has undesirable effects.
* During the preparation of this paper the author was a recipient of a Postgraduate Scholarship, Deutscher Akademischer Austauschdienst, Dortmund, KFA Abteilung Algenforschung, Germany. The author would like to thank L. Orl6ci, P. Juhgtsz-Nagy and E. van der Maarel for valuable suggestions, and comments.
success and group forming success in comparison to it. Campbell compared the so-called case 1 and case 2, in testing some measures, each containing two relev6s with conveniently compiled species abundances. This method of comparison has been well known for some time and the procedure (COREM) to be presented in this paper is a simple expansion of it. The resemblance values are calculated not for two, but for several cases ordered in series, which enables a graphical representation. Only quantitative measures are considered, i.e. measures based on species performance (cover %, abundance, biomass etc.) and not presence/absence.
The goals of the new method COREM (Comparison of Resemblance Measures) are: 1) to graphically depict the different similarity measures for Vegetatio 48, 47-59 (1981).