PiscesLogoSmallerStill Multivariate analysis

Top  Previous  Next

Multivariate analysis is used when the objective is to search for relationships between, or classify objects that are defined by, a number of attributes. Generally, we seek to show the relationship between sites (or samples) using the measured variables (eg species) as the attributes. Data sets can be large, for example marine benthic or forest beetle faunal studies can easily require analysis of a matrix of 100 samples (stations) by 350 species, and thus multivariate analysis requires a computer. If the objective is to assign objects to a number of discrete groups then cluster analysis should be considered. If there is no a priori reason to believe the objects will or could naturally fall into groups, then an ordination technique may be more suitable. Ordination assumes the objects form a continuum of variation and the objective is often to generate hypotheses about the environmental factor(s) that mould community structure.

 

There is a considerable literature on multivariate techniques. Useful texts for ecologists are Legendre & Legendre (1998), Digby & Kempton (1987) and Kent and Coker (1992).