ISSN:
1573-8868
Keywords:
classification
;
cluster analysis
;
mapping
;
multivariate analysis
;
pattern recognition
;
point-density analysis
;
geochemistry
;
ore-bearing districts
Source:
Springer Online Journal Archives 1860-2000
Topics:
Geosciences
,
Mathematics
Notes:
Abstract Large sets of rock compositions can be grouped by the analysis of corresponding point distributions in multidimensional space, where the separate dimensions represent the chemical variables. The point density around each composition is estimated in a small, multivariate, rectangular interval. Establishment of the identity of specimens contained in the interval, and systematic comparison of the point densities surrounding each of them, leads to the recognition of density maxima, which represent the statistical modes of the rock types present in the set. The remaining specimens by the same operations can be assigned to the groups formed. Experimental results are given for two sets, one comprising a wide variety of metasedimentary and metavolcanic rocks from the ore-bearing Skellefte district and adjacent parts of Västerbotten, Sweden, and the other representing a less variable population of basic rocks from Norrbotten, Sweden. Linear discriminant analysis and existing geological information indicate that the groups obtained are statistically and geologically valid.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF02312721
Permalink