ENBIS: European Network for Business and Industrial Statistics

Wine Characterization through Multivariate Statistics

22 September 2009, 11:40 – 12:10

Abstract

Submitted by

Ana Pereira

Authors

Ana C. Pereira (1,2) Marco S. Reis (1) Pedro M. Saraiva (1) José C. Marques (2)

Affiliation

(1) Department of Chemical Engineering, University of Coimbra; (2) Madeira Chemistry Research Centre, Department of Chemistry, University of Madeira

Abstract

Several studies have been carried out recently in order to differentiate and authenticate wines, encompassing numerous analytical methodologies that generate large amounts of experimental data presenting strong correlations among them. In this regard, Multivariate statistical techniques have become a powerful tool to address such modern measurement systems, as they allow one to explore and to take the most out of the large amounts of data available, in order to deepen the understanding about relevant phenomena going on during wine production and to improve wine quality assessment.
This work is focused in the production of Madeira wine, a Portuguese fortified wine, famous for its exceptional longevity and peculiar bouquet. Our main research goal is centered in the characterization of Madeira Wine at different ageing stages, for different types of Wine, following a wine flavour chromatography data collection step.
An exploratory data analysis was conducted using two different tools: biplots and contributions plots. The latter was found to be suitable for making comparisons between the importance of the variables under study in explaining a given trend identified in the Principal Components Analysis subspace. In order to take advantage of the maximum amount of information provided by the chromatography data sets, a new approach was developed and tested to complete the exploratory data analysis study, which consists of estimating variable contributions, by considering their intrinsic variability. In this way, it was possible to analyze which volatile compounds have statistically significant and/or similar contributions regarding the observed separation of wine samples from different groups, in the principal components space.
The results reached so far indicate that the combined use of GC/MS results, together with appropriate advanced multivariate statistical techniques, does allow us to come up with adequate procedures to identify different types of Madeira wines in terms of their aromatic characteristics and ageing time, as well as to interpret the differences observed.

Return to programme