Skip to main content
Bull. Spec. CORESTA Symposium, Kallithea, 1990, p. 98, APTS 03, ISSN.0525-6240

Evaluation of tobacco grading systems by multivariate analysis of their chemical quality parameters

CASTANO J.I.; VARGAS L.R.; PALACIO F.J.
Compania Colombiana de Tabaco S.A., Investigacion y Desarollo, Medellin, Colombia.
Multivariate methods are being widely used in analyzing large data sets where each object is characterized by several variables. Three multivariate statistical techniques were used in this work : discriminant analysis, principal components analysis and cluster analysis, in order to evaluate different types, grades and varieties of tobacco using the results of their chemical quality parameters such as reducing sugars, nicotine, total nitrogen, total volatiles bases, petroleum ether extract, ashes, chlorides and pH. Important information was obtained about the operativity and efficiency of tobacco grading systems and their deviations, when a discriminant analysis of chemical data was made from samples already graded according to conventional physical criteria. The discriminating capacity of the chemical variables and the great usefulness of the technique to differentiate amongst types and grades will be explained. Through Mahalanobis distances a system that allowed the establishment of the order of similarity between types and grades was deduced, thus enabling the search of compatible tobaccos with similar chemical properties, an important aspect for blending purposes. By principal components analysis it was found that the system of eight variables could be reduced up to four components which represented 90% of the total information. In such a way the result of the discriminatory analysis made with the first eight chemical variables was reproduced by 95%. A data file with samples of leaf and stem of flue-cured, burley and black tobaccos was analyzed through a non-hierarchical cluster method, showing the great capacity of this technique, to gather objects of a similar chemical nature in homogeneous groups; the technique itself being advantageous in the development of better grading criteria. The application of these techniques involving physical and organoleptical parameters in addition to the chemical ones, will provide a powerful tool for future work in this area.