Proposta de um Teste Monte Carlo para unidimensionalidade de painéis sensoriais

Abstract

A panel is said unidimensional when all the assessors score the products in a similar way. The level of unidimensionality reflects the quality of training the judges received. The Principal Component Analysis (PCA) is a multivariate technique used to reduce dimensionality. After determination of the principal components it can be tested if most of the total variation is explained only by the first dimension. In the literature there are tests to determine the optimal number of principal components that should be retained. Since most tests are general, ie, are only concerned with the amount of components to be retained, the objective of this study was to propose a new test to check the unidimensionality of sensory panels. The evaluation of type I error rates and power of the proposed test was performed using Monte Carlo simulation. This test was compared with existing ones, such as: asymptotic eigenvalues (AVa), parametric and nonparametric bootstrap tests, Monte Carlo test for unidimensionality (TMCU) and the Fujikoshi test. For comparison, the tests which were general have been translated into the context of unidimensionality of sensory panels. Under multivariate normality, the proposed test is recommended, since it has high power and controlled the type I error rate effectively, always close to or less than the nominal value.