Relationship between flavour descriptors and overall grading of analytical panels for virgin olive oil

An analytical panel is used to evaluate the quality of virgin olive oil. Two regression procedures—step‐wise multiple regression and regression on principal components—have been used to construct equations that relate profile to overall grading of 57 samples of virgin olive oil collected from all over Spain. Every equation has an adjusted R 2 > 0.94. Forty new samples evaluated by another panel were used to verify the models. Regression on the principal components correctly classified 87.5% of samples, against 95% classified by step‐wise multiple regression. Kurtosis, outliers, multicollinearity and singularity were also studied before building the regression models.