Initial investigation of the sensory profiling data of the yoghurts by ANOVA revealed that 22 sensory descriptors were significant (p 0.05) in describing differences between the types of yoghurts based on the experimental factors (Table 4). The non-significant attributes described the aroma of the yoghurts (O-sour and O-cheesewhey) and were excluded for further analysis. All the texture attributes, except T-flourychalky, were highly significant (p 0.001) in discriminating among the experimental factors and their interactions. The attributes describing the aftertaste of the yoghurts and the odour of cream (O-cream) were significant (p 0.05 and p 0.001, respectively) in assessing differences between the fat levels of the yoghurts. Significant differences among low-fat yoghurts manufactured with MWP at two protein levels were reported by the sensory attributes O-cabbage (p 0.001) and F-buttermilk (p 0.01).
ANOVA-PLSR was utilized to further investigate the relationship between the experimental design (X-variables) and the sensory
profiling data (Y-variables). Fig. 1A displays the correlation loadings plot of the sensory descriptors from the first two significant latent variables (LV) (explaining 64% and 13% of the variation in sensory data, respectively). Fig. 1B shows the distribution of the yoghurts tested. The grouping of the descriptors in Fig. 1A revealed that the first dimension was closely related to texture characteristics of the yoghurts. On the right side, the descriptors Creaminess, Viscosity (both manual and oral), T-body, T-meltdown, T-softness, M-resistance