The Role of Factor and Regression Analysis in the Interpretation of Geochemical Reconnaissance Data

Abstract

The interpretation of exploration oriented geochemical data frequently requires the recognition of subtle features related to mineralization, from the more obvious geochemical expressions of bedrock and surface environments. A number of previous investigations have indicated the potential of various computerized interpretational procedures as aids in identifying these features in geochemical data. The present investigation was concerned with the interpretation of multi-element data from a 750 mile2 {1942.5 km2) area of the Notre Dame Bay district of Newfoundland. The area is underlain by a series of Ordovician and Silurian sediments and volcanics and intrusives overlain by glacial deposits mostly composed of glacial till. Massive sulfide mineralization including the Whalesback and Gullbridge deposits occur within the Ordovician volcanics. R mode factor analysis was employed to establish the character and distribution of the principle metal associations related to bedrock and surface environment contributing to the overall data distribution. The factor scores were regressed against the individual metal concentrations of the elements composing the respective factors, the resulting residuals of the metals reflecting the component of metal related to some sources other than those reflected by the metal associations established by factor analysis. Anomalous areas of residual copper and zinc distributions indicate the areas of known sulfide mineralization more closely than the untreated metal distributions. On this basis, anomalous areas of residual copper and zinc, unrelated to known sulfide mineralization warrant further exploration investigation. It is therefore concluded that a combination of factor and regression analysis on multi-element data from the Notre Dame Bay district of Newfoundland serves to highlight subtle though significant features in multi-element data possibly related to mineralization that were not apparent from a consideration of the untreated data.

Geochemical signatures and mechanisms of trace elements dispersion in the area of the Vale das Gatas mine (Northern Portugal)

“…This facilitates data interpretation. R-mode analysis is particularly relevant in working out a process oriented interpretation (Saager and Sinclair, 1974;Closs and Nichol, 1975) while the use of the Qmode analysis, focusing on individuals rather than variables, is well suited to study sample sets where the individuals may be thought of as mixtures of end members (Hitchon et al, 1971;Nichol, 1973;Saager and Sinclair, 1974). Table 2 shows the results of Principal Component Analysis (PCA) using log data.…”

Interpretation of geochemical data from panned concentrates of wadi sediments using R-mode factor analysis, Jabal Habashi Quadrangle, sheet 26F, Kingdom of Saudi Arabia

“…More rigorous treatments of factor analysis can be found in Davis, 1973 andHowarth andSinding-Larson, 1983. Examples of factor analysis applied to geochemical exploration are presented by Saager andSinclair, 1974, andGloss andNichol, 1975. The data were analyzed by the multivariate procedure of R-mode factor analysis (Miesch, 1980, method 2) which is based on the correlation coefficients between the variables.…”

Deciphering factors controlling trace element distribution in the soils of Karaduvar industrial-agricultural area (Mersin, SE Turkey)

“…FA technique reduces a large number of variables to a minimum number of ''new'' variables called ''factors'' by linearly combining measurements made on a number of variables (Davis 1986). Kaiser's Varimax rotation is generally applied to the ''new'' variables in order to find factors that can be more easily explained in terms of natural or anthropogenic processes (Closs and Nichol 1975). This rotation is called varimax because the goal is to maximize the variance of the ''new'' variable, while minimizing the variance around the ''new '' variable (StatSoft, Inc. 1997).…”

“…In addition, there may be contributions from anthropogenic sources (i.e., agricultural and industrial) that can produce distinct geochemical differences compared to natural background. The factor solutions provide information on the following: (1) loading-the strength of a particular variable in a factor and takes values between ?1 and -1; (2) communality-the amount of the total variability of each variable explained in a given factor model, which is a value close to unity, and (3) eigenvalue-the amount of the total data variability explained by each model (Closs and Nichol 1975). If a FA is successful, the number of factors extracted from the dataset will be smaller than number of variables used, communalities will be close to unity, and the factors will be readily associated with particular sources or processes (Davis 1986).…”