Simulação de modelos multiníveis com regressões adaptativas robustas em condições experimentais de armazenamento de cafés especiais em ambientes refrigerados

Abstract

Numerous factors contribute to the quality of specialty coffees, including storage and refrigeration conditions. In this sense, we can conjecture that the results of sensory evaluations may be corrupted due to measurement errors, mainly when tasters are not trained, causing the appearance of observation outliers. With this motivation, this study aimed to propose simulation scenarios considering parametric values obtained by adjusting a multilevel model incorporating robust adaptive regressions to the presence of outliers in an actual experiment with processed and non-processed coffees stored at different periods and temperatures. In this context, computationally simulated scenarios, in which the errors in the response of the variable sensory score could be made at L = 5 and 10 units, were considered. The proposed methodology to infer simulated sensory scores in an experiment defined by the experimental conditions of storage and refrigerated environments was feasible because it contemplates robust characteristics in samples characterized with up to 30% of outlier observations in the evaluated simulation scenarios.