overall quality score was successfully well-predicted from the 12
attribute score data (R2
Y, Q2
Y ¼ 0.871, 0.869) using 2 latent variables
(1 predictive and 1 orthogonal component).
Subsequently, statistically significant attributes for overall
quality prediction were selected according to VIP which is a measure
of the importance of each X variable from the point of view of
the correlation to the Y variable (24). Since the average of squared
VIP scores equals 1, a greater than one rule is generally used as a
criterion for variable selection (25e28). In this study, 95% confi-
dence interval of the VIP was estimated by jack-knifing method and
the criterion value was set so that the lower-bound of the 95% jackknife
confidence interval did not encompass one. A previous report
concluded that the use of the jack-knife confidence interval