Introduction

Compared to the past, ever larger amounts of data are being collected in astronomy, and the rate will continue to accelerate in the next decades. It is therefore necessary to work on bigger samples if full advantage is to be taken of all accessible information. It is also necessary to derive as much information as possible from the diversity of the data, rather than restricting attention to subsets of it.

One way to work effectively on large samples is to apply, and if necessary to develop, suitable statistical methods. Multivariate data analysis methods are not intended to replace physical analysis: these should be seen as complementary, and statistical methods can effectively be used to run a rough preliminary investigation, to sort out ideas, to put a new (``objective'' or ``independent'') light on a problem, or to point out aspects which would not come out in a classical approach. Physical analysis is necessary subsequently to refine and interpret the results.

The multivariate methods implemented all operate on MIDAS tables. Such tables cross objects (e.g. ultraviolet spectra, spiral galaxies, or stellar chemical abundances) with variables or parameters.

Among widely-used multivariate methods :

1.

Principal Components Analysis  

2.

Hierarchical Cluster Analysis

3.

Non-Hierarchical Clustering, or Partitioning

4.

Minimal Spanning Tree        

5.

Discriminant Analysis  

6.

Correspondence Analysis  

We will look briefly at each of these in turn. Since comprehensive ancillary documentation is available (see references), we will not dwell on background material here. The wide-ranging applications for which multivariate statistical methods have been employed in astronomy also cannot be catalogued here (again, see the references given). In the following sections we will concentrate on various practical aspects of the methods. Note that the MVA context is required to activate the relevant commands (by using the command SET/CONTEXT MVA in a MIDAS session).