Visualizing multivariate analysis - An intuitive approach to high dimensional statistical extractions
AuthorsLewis, Stephen J.
AffiliationChester College of Higher Education
MetadataShow full item record
AbstractThe numerical output of multivariate statistical analyses may extend to a greater number of dimensions than can be comprehended and so may appear abstract and divorced from the original data. A need arises, therefore, for the provision of a more intuitive understanding of the results of such techniques - perhaps of a graphical nature. A simple method is to plot, what have come to be known as, Andrews' curves. A tabular procedure, using a standard computer spreadsheet, is described whereby the coefficients produced by various multivariate statistical techniques can be substituted into a simple equation to produce a smooth, wave-like curve characterising the source data. Importantly, this technique also provides a means whereby groups of curves may be compared visually to identify clusters and curves of similar or dissimilar overall shape. Similarly, "outliers" may also be spotted.
CitationIn K. Boyle & S. Anderson (Eds.), Computing and statistics in osteoarchaeology (pp. 31-34). Oxford: Oxbow Books, 1997.
DescriptionThis is the author's PDF version of an book chapter published in Computing and statistics in osteoarchaeology ©1997. The paper was originally delivered at the second meeting of the Osteoarchaeological Research Group at the Institute of Archaeology, University College, London on 8 April 1995.
The following license files are associated with this item: