confidence ellipses can
be constructed around treatment means to indicate 95% confidence
intervals (Heymann & Noble, 1989). Ellipses that do not touch indicate
that the treatments were significantly difference from one
another (p < 0.05) while taking into account the panelist means
and replicates. Significant canonical variates were determined by
Bartlett’s Test (Chatfield & Collins, 1980) and Scree plots (Chung,
Heymann, & Grün, 2003).