figure

In many cases, more complex statistical analyses are needed which deal with more than one variable at a time - these are called multivariate analyses. For instance, it may be necessary to divide the collecting sites into separate groups according the overall climate (e.g. mean annual and monthly rainfall, mean minimum/maximum temperature etc.), rather than use a single environmental variable such as rainfall.

A wide range of multivariate analysis techniques exist for analysing groups of ecogeographic or taxonomic variables. No single method will give predictive results in all situations - the actual method selected will depend on the aim of the analysis, the structure of the data, the availability of software and any advice given by a statistics expert. These analyses are only possible with robust data.

There are two main types of multivariate analysis of particular relevance to ecogeographic data - hierarchical clustering methods and ordination methods.

Hierarchical clustering

This divides objects into different groups, called clusters, on the basis of their similarity; two objects in the same cluster are more similar than two objects in different clusters. The objects may be plant populations, collecting sites or other type of sites. The clusters identified can be used as entities in further analysis e.g. distribution maps and regression analysis.

The clustering process may be of two types: (1) agglomerative - in which objects are joined into progressively larger clusters; or (2) divisive - in which the cluster of all objects is progressively divided into smaller clusters (not discussed here).


IPGRI Home Copyright © 1997, International Plant Genetic Resources Institute.