Multivariate Correlation between Color and Mineral Composition of Honeys and by Their Botanical Origin

• Published in: Journal of agricultural and food chemistry Vol. 53; no. 7; pp. 2574 - 2580
• Main Authors: González-Miret, Maria Lourdes, Terrab, Anass, Hernanz, Dolores, Fernández-Recamales, Maria Ángeles, Heredia, Francisco J
• Format: Journal Article
• Language: English
• Published: Washington, DC American Chemical Society 06-04-2005
• Subjects: Analysis of Variance; Biological and medical sciences; classification; Color; Confectionery products and chocolate industries, honey; discriminant analysis; equations; Flowers; food composition; Food industries; Fundamental and applied biological sciences. Psychology; honey; Honey - analysis; linear discriminant analysis; mineral content; Minerals - analysis; monofloral honey; multiple linear regression; multivariate analysis; Plants - chemistry; Regression Analysis; Origin; Honey; minerals; Discriminant analysis; Correlation; Linear regression; Color; Chemical composition; Inorganic element; multiple linear regression (MLR); Multiple regression; ICP-OES
• ISSN: 0021-8561; 1520-5118
• DOI: 10.1021/jf048207p

The mineral content and color characteristics of 77 honey samples were analyzed. Eighteen minerals were quantified for each honey. Multiple linear regression (MLR) was used to establish equations relating the colorimetric CIELAB coordinates to the mineral data. The results obtained shown that lightness (L*) was significantly correlated with S, Ca, Fe, As, Pb, and Cd for the dark honey types (avocado, heather, chestnut, and honeydew). For the light and brown honey types (citrus, rosemary, lavender, eucalyptus, and thyme), C ab* and b* showed the lower correlation with the mineral content of the honeys; their regression functions involve a few independent variables (Mg and Al for b* and only Al for C ab*). Furthermore, by means of application of linear discriminant analysis to the mineral content, it was possible to obtain a model that classifies the honeys by their lightness. The prediction ability of the built model, determined with the test set method, was 85%. Keywords: ICP−OES; minerals; honey; color; multiple linear regression (MLR); discriminant analysis