Multivariate statistical methods can be applied to analyse a complete set of multidimensional geochemical data and to identify latent relationships among these data. In this paper, we used multivariate statistical analysis, including k-means clustering, principal component analysis (PCA) and factor analysis (FA), to study the similarity of the sampling points and the relationships between metal mineralization and geological environment in the Nanling metallogenic belt, South China. The dataset consists of 1617 sediment samples analysed for 39 elements. The dataset was divided into three clusters by k-means clustering which were strongly associated with the distribution of lithostratigraphic units and the level of metal mineralization. Each cluster was analysed by PCA to identified principal components. Three factors extracted by the factor analysis explained c. 62% of the total variance and allow identification of the dominant ore-forming environment. Factor 1 describes c. 30% of the common variance and is highly loaded by Zn, Cu, Cr, Co, Ni, Mn, P, Ti, V, Mg and Fe. Factor 2 includes rare metals, rare earth elements (REE) and radioactive elements with Y, La, Nb, Zr and U, explaining c. 19% of the common variance. Factor 3 describes c. 14% of the common variance and is highly loaded by W, Sn, Mo, Pb, Be and Bi, representing tungsten polymetallic mineralization. In this paper, the Student’s t-test derived from weights-of-evidence modeling was used to measure the significance of spatial correlation between factor scores and mineral deposits.