The additive main effects and multiplicative interaction (AMMI) model is us
ed to investigate genotype x environment interactions often encountered in
variety trials. Interaction effects are accounted for by several multiplica
tive terms in the model, with each consisting of a genotypic and an environ
mental parameter. These effects are estimated using principal-component ana
lysis. An alternative method is presented to estimate the genotypic and env
ironmental parameters in the multiplicative terms of the AMMI model by cano
nical variate analysis. The number of significant interactive terms can be
easily determined using the chi(2) test. Each of the multiplicative terms i
nclude a unitless coefficient measuring genotypic response and an environme
ntal effect in terms of the physical unit of the concerned trait. Biplots c
an be used with the scores of both genotypes and environments for graphic e
xamination of analytic results. Total tuber-yield data from potato genotype
s tested in a series of international trials are used to illustrate the new
analytic procedure.