On objectives of methods of ordination

摘要: The objectives of a number methods ordination are examined and major distinction made between two approaches. first these has as primary objective the efficient redescription data, is typified by principal components analysis. However linear additive model implied in component analysis predominance unique variance, together with lack scale invariance suggests that other dimensionality reduction might be more appropriate ecologically—either non-metric multidimensional scaling or factor second approach, Curtis McIntosh continuum analysis, seeks to order stands so resulting data matrix particular form, not directly concerned reduction. Continuum only such pathseeking method, several others briefly examined. Finally Hill for seriation intrinsic approach Trunk seem provide close those required examination ecological data. Concluding comments on problems interpretation effects sampling description value results, especially light present tendency employ simulated test efficacy