{"title":"Canonical Correlation-based Tests for the Agreement of Sensory Panelists","authors":"M. C. Rocha, E. Ferreira, D. F. Ferreira","doi":"10.3844/JMSSP.2020.9.19","DOIUrl":null,"url":null,"abstract":"The reliability of the results of sensory analysis is directly linked to the performance of panel of assessors what, in general, means the ability of judges to identify small differences between products, the replicability of their ratings for the same product and the panel consonance. The panel consonance - usually called unidimensionality - can be understood as the agreement between the judges, thus it reflects the degree of training. Several methods have been proposed for assessing panel unidimensionality but always checking one attribute at a time. We proposed a generalization of the unidimensionality concept based on canonical correlation and enabling to consider several attributes simultaneously.","PeriodicalId":41981,"journal":{"name":"Jordan Journal of Mathematics and Statistics","volume":"81 1","pages":"9-19"},"PeriodicalIF":0.3000,"publicationDate":"2020-01-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Jordan Journal of Mathematics and Statistics","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.3844/JMSSP.2020.9.19","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q4","JCRName":"MATHEMATICS","Score":null,"Total":0}
引用次数: 0
Abstract
The reliability of the results of sensory analysis is directly linked to the performance of panel of assessors what, in general, means the ability of judges to identify small differences between products, the replicability of their ratings for the same product and the panel consonance. The panel consonance - usually called unidimensionality - can be understood as the agreement between the judges, thus it reflects the degree of training. Several methods have been proposed for assessing panel unidimensionality but always checking one attribute at a time. We proposed a generalization of the unidimensionality concept based on canonical correlation and enabling to consider several attributes simultaneously.