Characterization of european virgin olive oils using fatty acids

Se ha estudiado la posibilidad de discriminación entre aceites de oliva virgen procedentes de diferentes regiones de España, Italia y Portugal, mediante sus ácidos grasos. Para llevar a cabo la discriminación entre los aceites, desde su agrupación inicial hasta la diferenciación por regiones, se empleó una estructura arborescente. Fueron aplicados procedimientos estadísticos multivariantes para determinar la capacidad discriminante actual de este grupo de compuestos. La certeza asociada a la discriminación final fue calculada mediante la teoría de la evidencia.


INTRODUCTION
The methyl esters of fatty acids are considered to be of great use for the chemical characterization of olive oil (Derde et al., 1984;Forina et a!., 1983).Insufficiently checked data could artificially increase the discriminatory capacity of this series of compounds.Although, the methodology for their determination is standardized, the simultaneous use of data obtained from different columns or mixtures of oils of diverse qualities -virgin, pure or "lampantes" -is more often responsible for this unusual discriminatory capacity.When the oil is extra-virgin and the quantification methods identical for all samples, the discriminatory capacity diminishes markedly.
In the present work, the value of the fatty acids in discriminating between extra-virgin olive oils from various regions of Spain, Italy and Portugal have been determined (Cad. Oleícola, 1983).A dendrogram permitted differentiation between the oils starting from their initial grouping and progressing to differentiation on the basis of regions.Multivariant statistical analyses have been applied to determine the actual discriminatory capacity of this group of compounds.The confidence associated with the final differentiation was ascertained by means of the evidence theory (Aparicio, R., 1988), as opposed to other methods which use the probability theory (López Sabater et al., 1986) or expert discriminant systems that do not implement possibility theory (Derde et al., 1987).
Numerical and graphic representations demonstrate the discriminatory capacity of the fatty acids, using the probability theory in painA^ise comparisons made on the elements of each category.Given the complexity of their numerical representation, graphic representations on maps of Spain, Italy and Portugal have been used to show, by way of example, the results of this possibility theory.

Chemical methods
1349 samples of virgin olive oil were analyzed by means of gas chromatography, following norm UNE 55037, 1973.

Data analysis and statistical procedures
The results of tests of skew and kurtosis indicated that the data could be considered normal.No normalization function was applied on the data, as they satisfied the conditions of normality described by Tabachnick and Fidell (Tabachnick et al., 1983;Fisher, G.A., 1936).The usefulness of the F-to-enter test in reducing the possibility of the classificatory equations being due to chance was verified (Lavineetal., 1988).
The rules of decision were constructed on the basis of the discriminant equations obtained by means of the "Stepwise Discriminant Analysis" (SLDA) procedure, ¡nclu-ded in the SPSS+X package (2^ edition, 1986).The procedure was programmed with the following options: 1. F-to-enter > 4.96, taking into account a chi-squared distribution at 95.0%.
2. F-to-remove equal to the smallest value calculated by Wilk's Lambda.

Presentation and analysis of the results
The level of significance with which it is possible to affirm that the variances of two distributions are different, are presented in several tables.The following scale, expressed in linguistic terms, has been used: The expert system SEXIA has been used to deduce the classificatory final results, using the evidence theory (Aparicio, R., 1988;Aparicio et al., 1981a).An AUTOCAD graphic design package (1991) was used to portray the results on geographical maps of the countries on which the discrimination was performed.

RESULTS AND DISCUSSION
Figure 1 shows the dendrogram that was designed to determine the discnminant capacity of the fatty acids.Each level shows the groups to be distinguished using this senes of compounds.In the first level the differences between the oils from Spain, Italy and Portugal are analyzed.The discrimination was performed by means of pairwise comparisons of the countries e.g.Spain-Italy, Spain-Portugal, etc.

SEXIA
The following level analyzes the subgroup corresponding to the regions category.At the same time the discriminant capacity of the fatty acids in relation to the categories "provinces" and "zones" was studied.At these levels it was necessary to increase the number of chemical parameters to achieve better discriminatory results.

Discriminant analysis between countries
Figure 2 represents in computerized chromatograms (Aparicio et al., 1981b) the mean values of the seven fatty acids most representative of the oils from the three countries analyzed (table I).As an overall conclusion, it can be noted that there are differences in the fatty acids in the various olive oils, there being a tendency for the mediterranean olive oils to be distinct from the Portuguese oils.The predictive probability levels for countries differentiation are in the cases of Spain-Italy 88.91%, Italy-Portugal 98.77% and Spain-Portugal 98.52% of correct classification.It can be seen that, as the latitude of these countries is similar, the predictive levels increase gradually with longitude.The computerized chromatograms, together with the probability theory, also allows relational rules to be esta-

Figure 1
Figure 1The arborescent structure representing a specific finding taxonomy of SEXIA Expert System.