“This paper describes a new statistical method for the analysis of whole sets of variables simultaneously...”

This paper describes a novel methodology that allows multivariate data to be examined on the basis of any dissimilar measure of choice. I suspect that it is being cited in a number of different fields because its potential application is so broad. Virtually any set of multivariate data which is "misbehaving" statistically (such as species abundance data in ecology, for example) may be better analyzed through the use of an appropriate (non-Euclidean) distance function and randomization tests, using our proposed technique.

This paper describes a new statistical method for the analysis of whole sets of variables simultaneously (in response to some experimental treatments, for example). The method is unique and extremely useful because it can be applied even when traditional approaches fail. This commonly happens when variables have lots of zeros, are highly skewed, or when there are more variables than there are observation units. All of these problems occur, for example, in the analysis of species variables in whole ecosystems. This method provides an intuitive way of viewing information in a multivariate data cloud, and also allows rigorous tests of hypotheses from experiments.

I am originally a marine biologist and ecologist by training, who later became more and more involved in quantitative statistical analyses. The focus of my research now is to develop methods of statistical analysis that are appropriate for ecological and environmental data. I teamed up with Professor John Robinson (University of Sydney) initially, to get a more formal training in statistics, and also to work on the development of these kinds of methods (multivariate analysis and randomization tests). It just so happens that the techniques we have developed, which I was interested to apply to ecological data, are flexible enough to be used in virtually any context where variables do not fit the mold required by traditional statistical approaches.