Visualisation in multivariate and multidimensional data analysis

Visualisation may be used (a) in developing and presenting methodology and (b) in presenting the results of statistical analysis. Examples of both uses are discussed but mostly we are concerned with (b). Even within (b) we may distinguish between the basic mode of presentation and ancillary information needed to interpret diagrams correctly and which is rarely given by computer-produced visualisations. Information is needed on scaling of axes and geometrical concepts such as distance, inner-product, projections neighbourhood, angle, area, etc., and their interpretative value. Illustrations of basic notions will be driven by Procrustes analysis, multidimensional scaling, correspondence analysis, biplots, and asymmetry analysis. Both continuous and categorical values are considered. Some tentative proposals are made about labelling diagrams with information needed for their proper interpretation. The aim is to have self explanatory visualisations. The suggestions need further development and need to be extended to cover a much wider range of statistical disciplines.