Comparison of some techniques of multidimensional scaling in soil fauna ecology

The aim of this paper is to propose some alternative solutions to a problem proposed by El-Shishiny and Ghabbour, where the difficulty lies in vagueness rather than randomness. We first propose a city-block analysis, undoubtedly better adapted to this type of problem than a Euclidean analysis. In the second part we develop a technique consisting in the decomposition of every distance D_ij into two parts: the first, a star distance, considered as a personal effect; the second considered as a structure effect. These techniques are illustrated and compared in a real-life example.