Naslov
Multivariate distance and classification of winter wheat

Autori
Pecina, Marija ; Gunjača, Jerko

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, sažetak, znanstveni

Izvornik
Quantitative Genetics and Breeding Methods : The Way Ahead / - Pariz : EUCARPIA, 2000, 136-136

Skup
EUCARPIA - XIth Meeting of the Section Biometrics in Plant Breeding

Mjesto i datum
Pariz, Francuska, 30.08.2000. - 01.09.2000

Vrsta sudjelovanja
Poster

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
mahalanobis distance; MDS; CDA

Sažetak
Winter wheat selection in Croatia has a long tradition. The main breeding centers are near Zagreb and Osijek. It seems that the varieties selected in these centers have their own specific origin, partly responsible for their discrimination. The aim of this investigation was to analyze multivariate differences of the varieties classified by their selection programs. We chose four of them which had been present in official variety trials from 1994-96 with more than 10 varieties: Osijek Agricultural Institute (OS) and Agrigenetics (AG) from the Osijek area, and Botinec Institute (BC) and Faculty of Agriculture (ZG) from the Zagreb area, and put them together with three standard varieties (checks) as objects. A total of 22 variables, regularly measured in the testing process, were reduced to 16 in order to avoid the problems of singular covariance matrix. We used generalized squared Mahalanobis distance (D2M) as an input data matrix for comparison of the objects by multidimensional scaling (MDS), and canonical discriminant analysis (CDA) for classification of the varieties by selection programs. The varieties from the BC and the ZG programs were significantly different from all the others, while the difference between the AG and the OS selections was insignificant. The coordinates, DIM1 and DIM2, from MDS on D2M distances were used to construct the distance map of objects (programs and checks). Most of the detected distances and differences were confirmed by the map, except one: BC vs. ZG. The reason for that could be the presence of n-dimensional space represented by only two dimensions. The relative relationships between the objects can be established only by coordinates. First three CAN variables from CDA explained 83% of the total variance and the most important for discrimination of the selection program were sedimentation (r=0,464) and hectoliter weight (r=0,382) Nevertheless, ignoring all the other aspects, the distinctness of Croatian main winter wheat selection programs, based upon the multivariate data, was detected which in a certain way proved the existence of their characteristic germplasm.

Izvorni jezik
Engleski

Znanstvena područja
Poljoprivreda (agronomija)

POVEZANOST RADA