Case study of modeling covariance between external factors and sensory perception of coffee

Analysis and inference of sensory perceptions in coffee beverages are complex due to numerous random causes intrinsic to productivity, preparation, and especially consumer and/or taster subjectivity. In this context, latent variables often composed of a combination of other observed variables are discarded from conventional analyses. Following this argument, this study aimed to propose a model of structural equations applied to a database, geographical indication of coffees in Serra da Mantiqueira, with a methodological contribution characterized by inclusion of a treatment effect, contemplated by different altitudes at which coffees were produced. From the methodology used, a covariance structure was estimated, and used in another statistical methodology to discriminate the effects. It is concluded that the proposed model proved to be advantageous for allowing the analysis of the relationship of latent variables, production and environmental variations, which are not considered in a sensorial analysis, and showed that, in fact, they influence the sensorial perception, for the coffees produced in the Serra da Mantiqueira region. The correlation structure generated from the covariance matrix adjusted by the model resulted in estimates that could be used in other statistical methodologies more appropriate to discriminate the effects, exemplifying the use of principal components. Key words: Latent variable; adjusted goodness-of-fit (AGFI); altitude; goodness-of-fit (GFI).