6533b825fe1ef96bd1283354

RESEARCH PRODUCT

A matching task as a potential technique for descriptive profile validation

François SauvageotLynda O'neillSophie Nicklaus

subject

0303 health sciencesMatching (statistics)Nutrition and Dietetics030309 nutrition & dietetics04 agricultural and veterinary sciences[SDV.IDA] Life Sciences [q-bio]/Food engineering040401 food scienceSensory analysisRegressionCorrespondence analysisCanonical analysisTask (project management)03 medical and health sciences0404 agricultural biotechnologyDiscriminantConsistency (statistics)Statistics[SDV.IDA]Life Sciences [q-bio]/Food engineeringAlgorithmComputingMilieux_MISCELLANEOUSFood ScienceMathematics

description

If panellists can successfully match products to the corresponding descriptive profiles, then the profiles can be regarded as product-relevant and valid. This work examined the ability of a trained panel to perform a matching task between products and their descriptive profiles. A 13-member panel, trained to assess eight cheeses in terms of 19 flavour attributes, performed the task based on their individually developed profiles. The panel's ability to match products to profiles was well above that expected by chance, and chi-square statistics for each of the products were significant (P<0.05). A correspondence analysis based on the group results indicated that all the products were relatively close to their profiles, although there was some confusion amongst some of the products. The products that led to confusion were shown to be close to each other by a canonical analysis based on the profile data. Regression analyses indicated that measures of panellist discriminant capacity significantly predicted the frequency of correct matchings (P<0.05), whereas degree of consistency did not. Overall, at the group level, the matching task appeared to be a good measure of the product-relevance of the profiles.

https://hal.inrae.fr/hal-02679997