ABSTRACT: Hierarchical cluster (HCA) and canonical correlation (CCA) analyses were employed to explore the multivariate relationships among chemical components (proximate, mineral and lipidic components) of lean beef longissimus dorsii lumborum (LDL) and selected carcass traits of cattle fattened on pasture under tropical conditions (bulls, n = 60; steers, n = 60; from 2.5 to 4.0 years of age, estimated by dentition). The variables backfat thickness (BFT), Ca, Mn, Cu, C14:0, C15:0, and C20:0 showed the highest coefficients of variation. Three clusters were defined by the HCA. Out of all carcass traits, only BFT differed significantly (p < 0.001) among clusters. Clusters significantly (p < 0.001) differed for total lipids (TLIPIDS), moisture, dry matter (DM), fatty acid composition, cholesterol content, and mineral composition (except for Fe). The variables that define the canonical variate "CARCASS" were BFT and degree of marbling (MARBLING). TLIPIDS was the main variable for the "PROXIMATE" canonical variate, while C16:0 and C18:1c had the most relevant contribution to the "LIPIDS" canonical variate. BFT and MARBLING were highly cross-correlated with TLIPIDS which, in turn, was significantly affected by the IM lipid content. Carcass traits were poorly correlated with mineral content. These findings allow for the possibility to develop selection criteria based on BFT and/or marbling to sort carcasses, from grass-fed cattle fattened under tropical conditions, with differing nutritional values. Further analyses are needed to study the effects of sex condition on the associations among carcass traits and lipidic components.