Jump to Main Content
A Two-Stage, Three-Way Method for Classifying Genetic Resources in multiple Environments
- Jorge Franco, José Crossa, José Villaseñor, Alberto Castillo, Suketoshi Taba, Steve A. Eberhart
- Crop science 1999 v.39 no.1 pp. 259-267
- Zea mays, simulation models, agronomic traits, genotype-environment interaction, plant genetic resources, statistical analysis, gene banks, Mexico
- Cluster analysis is commonly used when gene bank accessions are evaluated in different environments and several continuous and categorical attributes are measured. The objectives of this study were (i) to develop a two-stage classification strategy for three-way data (accession × attribute × environment) using the Modified Location Model (MLM) with categorical and continuous variables and (ii) assess this approach for classifying observations of one simulated data set (with known structure) and two experimental data sets consisting of multi-attribute, multi-site field trials of Caribbean and Cónico accessions of maize (Zea mays L.). Clusters were compared including and excluding the categorical variables. Results from the simulated data showed good recovery of the group structure and a balanced effect of both categorical and continuous variables on the classification. The two-stage, three-way MLM method formed well-defined groups of accessions and characterized them based on the continuous as well as the categorical variables. As expected, groups formed using only continuous variables showed no clear response patterns concerning the categorical variables; however, those formed using both types of variables had clear response patterns with respect to each type. Furthermore, groups of Caribbean accessions showed dear association with geographical origin. The MLM model should be useful for classifying genetic resources evaluated in multi-environment trials into homogeneous groups with the objective of forming core subsets. When a large number of categorical variables are combined across environments, the MLM may not improve the initial classification obtained from hierarchical clustering strategy.