MuDiSc: Multi-Dimensional Scaling with Matlab and R

An increasing number of detrital zircon provenance studies are based on not just a few but many samples. This trend is likely to continue as the price of zircon U-Pb analyses continues to drop. The large datasets resulting from such studies call for a dimension-reducing technique such as Multi-Dimensional Scaling (MDS). Given a dissimilarity matrix (i.e., a table of pairwise distances), MDS constructs a 'map' on which 'similar' samples cluster closely together and 'dissimilar' samples plot far apart. This website presents some software tools for MDS analysis in the context of detrital geochronology, using the two-sample Kolmogorov-Smirnov statistic as a dissimilarity metric. Two alternative sets of tools are presented here, written in Matlab (Section 1) and R (Section 2). Further detail about these methods is provided in an accompanying paper:

Vermeesch, P., 2013, Multi-sample comparison of detrital age distributions. Chemical Geology, v.341, 140-146. doi:10.1016/j.chemgeo.2013.01.010.

1. A user-friendly Matlab-GUI:

Downloads: - MuDiSc.zip - a Matlab GUI to generate MDS and QQ plots of detrital age distributions. - DZages.xls - example input file for MuDiSc containing the U-Pb dataset from China.

2. MDS (and more!) with the provenance package in R:

The DZages.Rdata dataset needed to run the example code in the Chemical Geology paper can be downloaded from the bottom of this page. However, for practical purposes, it is more convenient to use the newly released provenance package. In addition to ordinary Multidimensional Scaling of detrital zircon age distributions, this package performs a host of other tasks as well, including MDS/PCA analysis of compositional data such as heavy mineral counts, Procrustes Analysis and 3-way MDS, and data visualisation as Kernel Density Estimates, Cumulative Age Distributions, pie charts and ternary diagrams. For further details about this comprehensive toolbox, see http://provenance.london-geochron.com.

Downloads: - DZages.Rdata - input data for the R code in the paper - DZages.csv - .csv-file with the same data - HMdata.csv - .csv-file with the heavy mineral data - BigData.Rdata - 5-proxy dataset from Namibia