Multivariate statistical techniques applied to five subsets of a multielemental lithogeochemical database to test and validate a geochemical typology classification model for the Paraná igneous province

Over the decades, authors have produced geochemical datasets from rock samples collected throughout the Paraná Igneous Province (PIP), and rock classification methods have used isolated or combined geochemical variables as criteria for classification. These data were compiled and standardized, creating a database containing information on geographical location and the concentration of multiple geochemical variables (elements or compounds) from outcrop and subsurface samples. In 2018, a new classification model for PIP rocks adopting geochemical types was proposed using the natural gaps in the frequency distribution curves and probability graphs of four discriminant variables: SiO2, Zr, TiO2, and P2O5. This paper presents the original database, the exploratory data analysis that allowed the extraction of five subsets, and the Multivariate Analysis techniques applied to test and verify the proposed classification model.