Application of the MARS modeling to estimate the gold values based on the paragenetic elements for an exploration data set

In the current investigation، the nonparametric modeling technique of multivariate adaptive regression spline، MARS، was used to reveal the underlying nonlinear patterns hidden in complex high dimensional data in shear zone gold mineralization-NW of Iran. MARS evaluated the elements of Pb-As-W-Bi-Cu-Be as paragentic group of elements. Different basis functions were introduced for illustration of the interaction between the response variable and independent variables، paragentic elements، according to the MARS method. The forward modeling and backward elimination steps were led to demonstration of the type and intensity of the relationship between independent variables and Au variable before and after each specific threshold value or knot. The comparison between estimated and actual values of the response (Au) demonstrated high correlation، showing that the response values were estimated by high accuracy. Different multivariate techniques should be combined if common statistical approaches are used، while this is the advantage of MARS technique that could be replaced instead of some multivariate common statistical techniques for the same estimation. The current investigation was done on secondary lithogeochemical environment where the effects of climate and physico-chemical conditions of the environment must be considered on the dispersion model.