U.S. flag

An official website of the United States government

PubAg

Main content area

Biplots of Linear-Bilinear Models for Studying Crossover Genotype × Environment Interaction

Author:
Crossa, Jose, Cornelius, Paul L., Yan, Weikai
Source:
Crop science 2002 v.42 no.2 pp. 619-633
ISSN:
0011-183X
Subject:
cultivars, genotype-environment interaction, models, regression analysis
Abstract:
Linear-bilinear models, such as the Shifted Multiplicative Model (SHMM) and Sites Regression Model (SREG), have been used to develop clustering procedures for finding subsets of sites (or cultivars) without cultivar crossover interaction (non-COI). Biplots of these models are useful for visual evaluation of cultivar responses across environments. The main purposes of this study were to investigate (i) SREG₂ and SHMM₂ biplots with the first multiplicative components constrained to be non-COI SREG₁ and SHMM₁ solutions, (ii) how the biplots can be used for identifying subsets of sites and cultivars with different levels of COI and with non-COI, and (iii) how these biplots compare with results obtained when clustering only sites or cultivars without cultivar rank change. Transformed and untransformed data from two multienvironment cultivar trials were used for illustration. Biplots from SHMM₂ and SREG₂ models graphically display the interaction variation due to low level COI or non-COI (first multiplicative term) versus the interaction variation due to COI (second multiplicative term). The biplots obtained by means of the non-COI first term constrained solution of the SREG₂ and SHMM₂ models have the same interpretability properties as the standard biplots obtained by means of the unconstrained solution. With the unconstrained and constrained solutions, it is possible to identify subsets of sites and cultivars with low level COI and non-COI. Biplots based on unscaled or scaled data produced similar results. Groups of sites and cultivars with low level COI and non-COI were similar to those found when only sites (or cultivars) were clustered into non-COI groups using the SHMM and SREG clustering approach.
Agid:
6655266