Statistical methods for classifying genotypes

In genetic resource conservation and plant breeding, multivariate data on continuous and categorical traits are collected with the objective of selecting genotypes and accessions that best represent the entire population or gene collection with the minimum loss of genetic diversity. Therefore, the best numerical classification strategy is the one that produces the most compact and well-separated groups, that is, minimum variability within each group and maximum variability among groups. In this study, we review geometric classification techniques as well as statistical models based on mixed distribution models. The two-stage sequential clustering strategy uses all variables, continuous and categorical, and it tends to form more homogeneous groups of individuals than other clustering strategies. The sequential clustering strategy can be applied to three-way data comprising genotypes x environments x attributes. This approach groups genotypes with consistent responses for most of the continuous and categorical traits across environments.