The effect of the type of MWP on the particle size data and the
level of native-to-denatured whey protein of the powders was
analysed by one-way ANOVA and Least Significance Difference
(LSD) as post-hoc test (PROC GLM, SAS version 9.1).
An analysis of variance followed by an LSD post-hoc test was
performed on the rheological data to assess differences amongst
the yoghurt types.
The sensory raw data was analyzed with regard to panel
performance using mixed model ANOVA, on individual descriptors
with products (n ¼ 23) and repetitions as fixed factors, and panellists
(n ¼ 11) and serving order of samples as random factors. This
resulted in the elimination of one panellist for some specific
sensory attributes.
The overall effect of experimental design treatments was
assessed by performing a new mixed model ANOVA on individual
descriptors using products and replications as fixed factors and
panellists as random factors. Least Significance Difference (LSD) was used as post-hoc test to compare between specific treatment
factors and their interactions (PROC MIXED, SAS version 9.1).
Furthermore, ANOVA-PLSR (APLSR) was applied to investigate
relationships between sensory data and the experimental design as
a graphical alternative to ANOVA (Martens & Martens, 2001).
Investigation of underlying relationship between rheological
and sensory data was done using Partial Least Squares Regression
(PLS2) (PROC PLS, SAS version 9.1). Prior to multivariate analysis,
the data was level corrected for scale differences and averaged over
assessors. Moreover, cross-validation was performed, leaving one
replicate out at a time. The quality of the regression models was
evaluated by the coefficient of determination (R2). Finally, linear
regression was performed between specific sensory attributes and
rheological variables to assess correlation and predictions (PROC
REG, SAS version 9.1).