Final taste test
To compare liking and perceived saltiness for each of the four
tomato juice samples between initial and final taste tests among salt
reduction groups, among the three motivation levels within each
group, and among the three hedonic sensitivity levels within each
group, we used mixed models with random effects for subject and
additional autoregressive covariance structure among the four sam-
ples tasted at one time, using the Kenward–Rogers degrees of free-
dom approximation (the autoregressive covariance structure
allows for observations that are further apart to be less strongly cor-
related). Liking and perceived saltiness intensity were dependent
variables in the models. Salt reduction group, tomato juice sample,
motivation level, and hedonic sensitivity level were predictors.
Interaction terms for salt reduction group, tomato juice sample,
and motivation level; and for salt reduction group, tomato juice
sample, and hedonic sensitivity level were also predictors. Salt
reduction group was coded using time-varying group membership,
so all individuals had the same group membership at the initial time
point, as the treatment had not yet been applied, but different group
memberships after treatment (Weiss, 2005). A post hoc Bonferroni
correction for multiple comparisons was used to test significant dif-
ferences between least squares means from the model. Linear
regression was used to determine the relationship between per-
ceived PROP intensity and overall liking and saltiness.
To determine if preference for salt in tomato juice changed
between taste tests for subjects in either salt reduction group, a
McNemar’s test was performed by determining the most preferred
sample for each subject served at the initial and the final taste test,
and then determining the number of subjects within each salt reduc-
tion group that either had no change or a shift in preference between
taste tests. A McNemar’s test statistic was determined and tested for
significance to determine if the hypothesis of no change in prefer-
ence for either salt reduction group could be rejected.