کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
1183935 | 1492084 | 2016 | 11 صفحه PDF | دانلود رایگان |
• Olive oil blends’ compositions were chemometrically predicted.
• Multivariate calibration method based on a mixture design was developed.
• Proportions of mixed olive oil cultivars were predicted from blends’ FA profiles.
• Prediction error analysis helped to identify the best predicted blends (errors < 5%).
• Mapping of errors in simplex space gave trajectories for high traceability blends.
Olive oil blends (OOBs) are complex matrices combining different cultivars at variable proportions. Although qualitative determinations of OOBs have been subjected to several chemometric works, quantitative evaluations of their contents remain poorly developed because of traceability difficulties concerning co-occurring cultivars. Around this question, we recently published an original simplex approach helping to develop predictive models of the proportions of co-occurring cultivars from chemical profiles of resulting blends (Semmar & Artaud, 2015). Beyond predictive model construction and validation, this paper presents an extension based on prediction errors’ analysis to statistically define the blends with the highest predictability among all the possible ones that can be made by mixing cultivars at different proportions. This provides an interesting way to identify a priori labeled commercial products with potentially high traceability taking into account the natural chemical variability of different constitutive cultivars.
Journal: Food Chemistry - Volume 208, 1 October 2016, Pages 150–160