The effect of panel selection and training on external preference mapping using a low number of samples

External preference mapping (PREFMAP) relates the preference data of a group of consumers to a multi-dimensional representation of the stimuli. This representation can be obtained from descriptive sensory data generated by a panel. Four vegetable oil samples were analysed. Three different stimulus spaces were obtained from generalized Procrustes analysis (GPA) of descriptive sensory data generated by three panels (12 assessors): a non-selected panel, a selected panel and the latter after training. The preference data of each consumer was regressed against the product co-ordinates obtained from each stimulus space using the vector model. The interpretation of stimulus space was easier with selected assessors, but in an overall sense the three stimulus spaces were similar. However, the slight differences between the different sensory spaces produced important differences in the PREFMAP interpretation.