Statistisk beskrivelse av geokjemiske data.

Abstract

A manual procedure for the organization of geochemical observations into frequency distributions is described. When frequency distributions of geochemical exploration data are plotted on logarithmic probability paper, more or less straight lines are often obtained, because trace element concentrations are in many cases approximately lognormal. Values for selected parametres for mean and deviation can be estimated from such plots. If the distribution plot is nonlinear, a mixture of two or more distributions may be indicated. If the mixture is bimodal, the empirical frequency distribution can be divided graphically into two theoretical distributions: two backgrounds, or background and anomaly. Often the background can simply be represented by a straight line through the lower and middle part of the distribution plot. When grouping the metal contents for map presentation, the geometric deviation of the background can conveniently be used as a unit for concentration intervals. A universal grouping can be obtained by dividing each tenth power of concentration into equal intervals on a logarithmic scale. The analytical sensitivity should be taken into account when setting the lowest interval limit, and the total data variability as well as the sampling and analytical error when deciding how many intervals each tenth power should be divided in.