报告题目：Many-Facet Rasch models: A key to hedonic measurement

主 讲 人（一）： Peter Ho  

报告题目：Examination of the acceptability of beef lasagne products in a consumer benchmarking test with a Many-Facet Rasch model

主 讲 人（二）： Zheng Li     

报告时间： 2018年4月25日（周三）上午9:00

报告地点：下沙校区食品学院257报告厅

报告人简介：

      Peter Ho is a Lecturer in Food Processing at the University of Leeds in the United Kingdom. His primary research interests at the School of Food Science and Nutrition are in Sensory Science, Food Choice and Measurement. He is the leader of the Sensory Measurement Analytics Research and Teaching Laboratory. His current research at SMARTLab has been focussed on:

1.Developing objective unidimensional measures for sensory acceptability measurement;

2.Techniques for improved food product development and optimisation;

3.Developing a universally calibrated person index for degree of chili hotness;

4.Applying a single satiety index for repeated studies on measuring satiety and satiation

He has published over 30 publications and is editor of Case Studies in Food Safety and Environmental Health and co-editor of Experiments in Unit Operations and Processing of Foods.    

报告摘要（一）：

      The 9-point hedonic scale is used widely for measuring overall acceptability. Parametric statistical methods, which require interval-scaled data, are often used to analyse raw data without considering the scale properties and how assessors interpret the meaning of category labels. Data from 9-point hedonic scale is ordinal. The numerical distance between each category label is assumed to be equal, by assigning numbers from 1 to 9 to the category labels, however this assumption has been proven to be invalid.  

      Rasch modelling provides a new approach to handling category scales, such as the 9-point hedonic scale, by firstly converting ordinal data into interval-scaled data, prior to analysing the transform data with parametric statistical methods. A single measure of overall acceptability was constructed using a Many-Facet Rasch model, from hedonic ratings of 10 individual attributes that were used by 90 panellists in comparing differences in the acceptability of smoked hams. The model assumption that all attributes fitted on a single dimension was confirmed by a principal component analysis of Rasch model residuals.

A single unidimensional plot represented the degree of acceptability of each product, the degree of satisfaction of each consumer and the relationship between each sensory attribute to the unidimensional measure of overall acceptability. The Many-Facet Wright map showed that the distance between category labels were unequally spaced, becoming increasingly larger the further away from the central category label. The Rasch measure of overall acceptability, that included the ratings of 10 attributes, was able to detect differences between the acceptability of the smoked hams that was not possible from single ratings of overall acceptability using the untransformed scores.

Many-Facet Rasch modelling provides a framework for developing overall interval-scaled sensory measures that can be represented on a single unidimensional continuum, that only requires panellists to use simple category scales.

报告摘要（二）：

This study explores the application of Many-Facet Rasch (MFR) models for examining the acceptability of 6 commercial lasagne ready meal products. A modified repertory grid method was used, whereby 45 participants generated a list of 20 attributes that described the aroma, appearance, taste/flavour and texture/mouthfeel characteristics. Each participant was given a pair of samples to compare similarities and differences and, through a 1-to-1 interview, required to select descriptions that best described what they perceived. The acceptability of all 6 lasagne products were subsequently rated by 96 panellists, based on the selected list, using the 9-point hedonic scale (from “Dislike extremely” to “Like extremely”).

Data collected was fitted to a MFR model to obtain a single measure of overall acceptability, using the Rating Scale model to model the Panellist, Product and Attribute facets. A dimensionality test, using a principal component analysis on Rasch model residuals, indicated that all individual attributes fitted adequately to a single dimension. All three facets were then compared using a Many-Facet variable map (figure 1) that allows all three facets to be placed on the same logit scale. The results showed that (1) two healthy-eating products were the least preferred products; (2) The panellists’ satisfaction levels distributed in a large range from “Dislike moderately” to “Like moderately”; (3) The “Visibility of vegetable chunks” had the lowest contribution to the overall acceptability measure used by the panellists. The information provided from the results can be used for the idea generation task of the development of new lasagne product.