Similarity analysis using rank in till geochemistry

MÄKINEN, JARI, 1991: Similarity analysis using rank in till geochemistry. Bull. Geol. Soc. Finland 63, Part 1, 49—57 A similarity analysis method based on Spearman's rank correlation coefficient (p) was developed aiming to indirectly measure the similarity between the geochemistry of till and bedrock. The study area lies in Central Finland, comprising c. 4800 km, and it includes Pyhäsalmi Säviä zone characterized by copper-zinc-mineralizations. The data consists of 1099 till samples with 11 variables; Al, Ba, Co, Cr, Cu, Li, Mn, Sc, Ti, V and Zn. Subset of 83 cases, representing Kiuruvesi schist belt area, was selected from the till data for the model. The model was formed from the median values of the subset. Model based on the till data consists of information on glacial and bedrock geology, and usage of different similarity measures consequently emphasize one or the other geological factors in the till. The till data covering the whole study area was compared to the model with an aid of similarity analysis, using the Euclidean distance, the cosine theta and the rank correlation coefficient as a measure. The results indicate that Euclidean distance reflects mainly information connected to the glacial geology and similarity pattern produced by cosine theta or rank correlation coefficient correlates very well with major bedrock structures. Similarity analysis of the till data based on rank reflected the bedrock structures better than using cosine theta as a measure. The rank correlation coefficient is nonparametric and consequently less insensitive to anomalous element concentrations than the common parametric cosine theta measure. Hence the rank correlation is readily applicable to geochemical data, which invariably include anomalous values.