De l'optimisation du nombre de variables morphologiques dans la discrimination entre populations (le cas de l'aile chez Drosophila melanogaster)

Whilst wing size measurements are often used in Drosophila to discriminate between groups of individuals, the choice of the particular characters measured is generally not justified.Statistical and genetical properties of 8 characters measured on the left wing, in three stocks of Drosophila melanogaster were analysed. A multivariate analysis is done by studying correlations between characters and by interpreting the first factor of a principal component analysis.The results shows that only one measurement is sufficient in order to discriminate between the stocks, as far as mean values are concerned; wing length is, in that case, the best character since its variance has a relatively high genetic component. When dealing with correlation among characters as a way of estimating shape, four measurements are needed and sufficient. This shows the necessity of choosing the characters to be measured.