Multivariate Extension of Griffing's Diallel Analysis

Diallel-cross designs have been used extensively in plant and animal breeding programs. In most cases, multiple traits are measured in the study, but the analyses are performed on each trait separately. The objective of this paper is to provide an alternative multivariate procedure so that the combining abilities among traits can be evaluated simultaneously and the interaction of combining ability and traits can be tested statistically.
Numerical example is provided from data of a rice study. Three characters, panicles per plant, percent sterility, and 100 kernel weight were used in the analysis. Results showed that some parental lines with greater combining abilities in one trait may not be so good in another trait. Also certain crosses may be better for one trait but not necessarily as good for other traits.