Visual clustering and classification: The Oronsay particle size dataset revisited (Q1297895)

In this article abilities of the software systems Xgobi, ExplorN and Manet for the interactive visual analysis of multidimensional data are compared. The authors analyse data on the sand from Oronsay island in the Inner Hebrides. The samples of sand were taken from ``modern dunes'', ``modern beach lines'' and from archeological sites of the Mesolithic period. The question is if the archeological samples are of ``dune'' or of ``beach'' type. The answer is connected to the problem of the seaward shift of the beach-dune interface. The distribution of particle sizes was measured for all samples, so 12-dimensional data vectors were obtained for 226 samples. Scatterplot rotation with projection pursuit, grand tour, parallel coordinate plots and linked low-dimensional views with user specified subsetting are used to investigate the cluster structure of the data. The authors' conclusion is that ``All three programs, Xgobi, ExplorN and Manet, and the related paradigms, have unique advantages and disadvantages that highly relate to the users experience and personal preferences. For the Oronsay particle size data we did not notice that any of the programs provided unequivocally better results''.