| Popis: |
Human assessors are currently the best instruments for evaluating the sensory properties of both food and non-food consumer products. In the early stages of descriptive sensory panel training, a lexicon is still undergoing refinement and assessors are only beginning to understand the identities of sensory attributes and how to use the scale to communicate the intensity of sensations. To expedite training, the panel leader provides structure and feedback. Several statistical approaches have been proposed that indicate training status. In this thesis several approaches are reviewed, and a multistage process that involves scrubbing data, estimating parameters, reviewing fit, and exploring multivariate relationships is developed. The process reveals attribute misunderstandings, considers candidate training targets for subsequent training sessions, and explores data to better understand product properties. |
