Multivariate statistical methods have become commonplace in the Earth Sciences. What was once an exclusive area of activity is now within the reach of Everyman, owing to the ubiquitousness of mini-computers and the ready availability of software for doing the computing. In the days when one was required to do one's own programming, it was necessary to acquire considerable proficiency in linear algebra and one or more programming languages. Today, the vast majority of the people who use multivariate methods to analyse geological data have little or no idea of the matrix operations underlying a particular method, nor, for that matter, what the program is actually supposed to be doing. This situation can be both good and bad. It can do no harm if everything goes according to schedule, the program being used is competently constructed, which, alas, is far from being the general case, and there are no strong deviations from standard statistical theory in the data under examination. It is bad if the data do not fit the theoretical requirements of a particular method and even worse if the method of computation used is inappropriate. It is an inescapable and sad fact of life that much geological and biological material deviates in some manner or other from the theoretical requirements of a multivariate statistical procedure. The immediate relevance of this observation is that there are many sources of error in doing an analysis of geological data by means of standard statistical software.
