Chemistry of Brazilian green coffee and the quality of the beverage, 5: Multiple linear regression analysis

Abstract

Multiple linear regression analysis was performed to try to relate the quality of the coffee beverage to its chemical composition. Six samples of each beverage quality (Soft, Hard, Riada and Rio) were classified by professional tasters from the I B.C., and 24 cup tests were performed for each sample. The chemical analyses of total free carbohydrates, reducing sugars, total nitrogen, proteins soluble in NaCl, NaOH and phosphate buffer, total chlorogenic acid, phenols soluble in methanol and water, hydrolyzable phenols and electrophotogenic behavior of the proteins were considered in the regression analysis.

When total nitrogen was omitted, an equation was obtained in which significance was obtained at the 5 percent level. The regression coefficient was 0.83. The coefficient of determination was 0.6943, meaning that 69.13 percent of the variation in beverage quality depends on the compounds analyzed, and can be explained by the multiple linear regression equation calculated in this work. Among the compounds studied, those that most affected beverage quality were total chlorogenic acids (negatively) and water-soluble phenols (positively).