Análise generalizada de procrustes via R: uma aplicação em laticínios

Abstract

In Brazil, for food evaluation data analysis it's common to use just univariate Analysis of Variance. This way, one looses the between variable information. To solve this problem one may use the multivariate ANOVA. The Generalized Procmstes Analysis (GPA), often performed in expensive softwares, is a multivariate exploratory data analysis with the aims: (i) remove scores' bias; (ii) resume the assessore' agreement to a plan plot; (iii) show the objects relative differences. GPA's geometrical and algebraic interpretations were given. Its implementation algorithm and Analysis of Variance were showed too. To illustrate its application, gorgonzola cheese data, from the south ofMinas Gerais state, were analyzed by uni e multivariate ANOVAs and GPA, on the free statistical software R. Ten attribute were evaluated. Experiment was conducted on a random blocs design (RBD) and analyzed in a 2 x 4 factorial scheme (two technology leveis: greater and smaller milk fat content and four ripen ages: 30, 45, 60 and 70 days). ANOVAs detected significative differences between technologies, ages and interactions. Generalized Procmstes Analysis provided the consensus configuration on a plan and showed relative differences among the cheeses and the both technologies' sensory evolution. The assessore considered higher milk fat content technology the best. GPA can be reasonably performed by free software, R, and should be utilized like a complement of Analysis of Variance.