In this work, the use of 1 H-NMR spectroscopy and a statistical approach to the analysis of biodiesel concentrations in blends with conventional diesel is described. For this, we performed 1 H-NMR analyses using distinct mixtures of biodiesel from soybean and castor oil in mineral diesel, in concentrations ranging from 0.5 to 30%, and then we applied partial least squares regression (PLS) and principal components regression (PCR) to such data. So, six models were designed and they were evaluated through statistical parameters and through the analysis of four samples prepared in the laboratory. Briefly, a PLS model, obtained through the selection of aromatic, aliphatic and methoxy spectral regions, was quite suitable for the prediction of biodiesel concentrations greater than 2.0%. Deviations of real and predicted values were found to B2 commercial blends, indicating that this model can only be applied to blends exceeding a 2.0% level of biodiesel in petroleum diesel. In conclusion, the 1 H-NMR-PLS method is fairly useful for the quality control of biodiesel-diesel blends, whose commercialisation has increased in the last few years.