Biometrical Letters vol. 46(1), 2009, pp. 55-70


Show full-size cover
USING MULTIDIMENSIONAL UNFOLDING IN PLANT BREEDING PROGRAMS

A.A.A.F. Martins1, I.M.S. Pinto1, M.G.M.S. Cardoso2

1Instituto Superior de Engenharia de Lisboa, Mathematics Cientific Area, Portugal, e-mail: anamartins@deea.isel.ipl.pt, ipinto@deetc.isel.ipl.pt
2ISCTE Business School, Departament of Quantitative Methods, Portugal, e-mail: margarida.cardoso@iscte.pt


In the present work Multidimensional Unfolding (MDU) is suggested as a new approach to support decision-making in plant breeding programs. It is an exploratory data analysis technique that yields the construction of a map, picturing the attractiveness of cultivars towards planting environments. The proposed approach is illustrated using data from a wheat plant breeding program in Portugal (1986-1999). MDU precedes the use of Linear Joint Regression Analysis (LJRA) (e.g. Pinto 2006) in the study of relationships between cultivars and environments, taking into account the cultivars' performance evaluation. As regards the selection of the best cultivars, both approaches agree. Furthermore, MDU provides an additional advantage related to its easily interpreted results. In fact simplicity and interpretability may be considered the main advantages of the proposed approach.


Multidimensional Unfolding, Linear Joint Regression Analysis, Plant Breeding Programs, Cultivar Selection.