RESULTS AND BIOLOGICAL INTERPRETATION OF SHIFTED MULTIPLICATIVE MODELCLUSTERING OF DURUM-WHEAT CULTIVARS AND TEST SITE

The examination of crossover interactions (COI) (genotypic rank change ) and identification of subsets of homogeneous groups of sites and cul tivars without COI are important in the interpretation of cultivar tri als in breeding and agronomy. The shifted multiplicative model (SHMM) clustering method was applied to a CIMMYT durum wheat (Triticum turgid um L. var. durum) yield trial consisting of 24 cultivars grown at 40 i nternational sites during the 1990-1991 season. The objectives were to identify subsets of sites and cultivars with negligible genotypic ran k change and attempt to give a biological interpretation for the resul ting clusters. The SHMM with one multiplicative term (SHMM(1)) provide d an adequate fit for seven final groups of cultivars and eight final groups of sites with small numbers of COI. Grouping of cultivars great ly reflected similarity in genetic backgrounds and, consequently, simi larity of response to test environments. Differential responses within the same genetic background were attributed to differences in simply inherited traits such as plant height and disease resistance. The obse rved grouping of sites was generally associated with latitude, while e nvironmental conditions that influenced crop phenology and cropping cy cle delineated the final groupings. The results suggest that groups fo rmed based on SHMM clustering methods have valid biological basis. Rou tine use of SHMM clustering methods could increase selection efficienc y through the identification and selection of superior cultivars withi n clusters having negligible COI. Similarly, groups of test sites that represent similar selection environments could be identified and that would facilitate identification of key test sites as well as decision making concerning exchange of germplasm and information.