Desenvolvimento e avaliação do desempenho de uma "Escala hedônica com referência" para utilização em teste de consumidor

AUTOR(ES)
DATA DE PUBLICAÇÃO

1999

RESUMO

In the last decade, the global economy promoted a high competitiveness amongst products that dispute the same consumer market, and as a consequence, brought an increase in the necessity for food industries to compare several formulations or processes with respect to a market leader through sensorial tests with consumers. The 9-point structured hedonic scale, the most commonly used scale in acceptability tests, evaluates the degree with which the samples are accepted by the consumers. However, it does not allow for a direct comparison between two samples, not putting the consumers in a real situation of choice between two or more samples. The present work concentrated on the adaptation of the traditional hedonic scale for use in consumer tests when the use of a reference sample is desirable. This adapted scale was called by us an "hedonic scale with reference" and its performance was evaluated through two experiments. In the first experiment, the performance of the hedonic scale with reference was compared to that of the 9-point structured hedonic scale and that of the magnitude estimation scale, with respect to the discriminative power amongst samples, with respect to the variability of the answers and to the adequacy of the answers to the necessary assumptions for the accomplishment of the analysis of variance (ANOVA). The number of significant discriminations (p ≤ 0.05) obtained between the samples and the reference sample was used as a criteria to judge the discriminative power of the scales and they were verified using the Tukey and Dunnett tests, as well as the significance levels of Fsample obtained in ANOVA and in "bootstrap" resampling simulations. The comparison amongst the variability of the consumer responses obtained with the scales studied was accomplished by applying the Tukey test to the mean variation coefficients of the scales The normality of the response distribution obtained with all scales was evaluated through the normal probability and box-whisker graphic constructions and analyses, and the Kolmogorov-Smirnov normality test. To evaluate the homogeneity of the sample variances, the correlation (pr) significance between standard deviations and acceptance means was tested and the Levene homogeneity test used. Although the magnitude estimation scale has presented greater discriminative power than the other scales, the results showed that its answers presented a greater deviation from normality and non significant homogeneous variances (p ≤ 0.05), which made the use of ANOVA inadequate to analyse the data collected with that scale. When the hedonic scale with reference were compared to the scale traditional hedonic scale, it presented smaller p values, showing larger discriminative power than the traditional hedonic scale. The mean variation coefficients of these two scales did not differ significantly (p ≤ 0.05), suggesting that the the consumer response variability was similar in both scales. The responses from the hedonic scale with reference were closer to the normal distribution and showed greater variance homogeneity than those obtained from the traditional hedonic scale. Thus, besides offering greater discriminative power, these results allow for the conclusion that the hedonic scale with reference supplied data which better fitted the assumptions of ANOVA, giving greater reliability to the results obtained. In the second experiment, the performance of the hedonic scale with reference was compared to that of the 9-point traditional structured hedonic scale in obtaining acceptability predictive models from consumer responses. Eleven formulations of reconstituted orange juice were prepared using a 22 factorial design called the central composite design. The independent variables studied were: °Brix (9.0, 9.7, 11.5, 13.2, 14.0) and acidity (0.68, 0.74, 0.90, 1.05, 1.12 g/100 mL). The global acceptance of the samples was evaluated by 44 consumers using the following two scales: a 9-point structured hedonic scale and a hedonic scale with reference. The data were analysed using Response Surface Methodology, analysis of the model residues and Dunnett s means test. Both scales generated predictive models of acceptance with high coefficients of determination (R2), these being: 0.89 (p = 0.001) for the traditional hedonic scale and 0.91 (p = 0.0005) for the hedonic scale with reference. The models generated by the two scales were similar with respect to: the non-significance of the lack of fit at p ≤ 0.05, the parameters comprising the final adjusted model and the adequacy of the residues to the statistical assumptions inherent in the regression analysis. The optimized regions were similar for the two scales. Dunnett s means test indicated that the hedonic scale with reference provided a greater discrimination of the samples with respect to the standard sample. The performance of the two scales in generating predictive models of acceptance was equivalent.

ASSUNTO(S)

consumidores avaliação sensorial

Documentos Relacionados