Performance of three affective methods and diagnosis of the ANOVA model

Consumer data arising from traditional scales frequently do not comply with ANOVA assumptions of normality, independence and homoscedasticity. Differential expansiveness of scoring between assessors is also a limitation in the use of scales. The self-adjusting scale was designed to minimize some of these problems but its performance with consumers has not been reported. This study compared the performance of the 9-point hedonic scale, the self-adjusting scale and the ranking test with respect to: differential expansiveness between assessors, discriminating power and compliance of data with ANOVA assumptions. Candies were evaluated by 288 consumers. Data normality was checked by normal probability plots, Kolmogorov¯Smirnov test and coefficients of skewness and kurtosis. Homoscedasticity assumptions were evaluated by the Levene test and scatterplots. The self-adjusting scale was effective to deal with differential assessor expansiveness, produced homogeneous variances but its data showed gross deviations from normality. Hedonic scale data showed slight deviation from normality and lack of homoscedasticity. Rank and self-adjusting scale showed the highest and the lowest discriminating power, respectively.