Biplot Analysis of Genotype × Environment Interaction: Proceed with Caution

Biplot analysis has been used for studying genotype × environment interaction (GE) or any two-way table. Its descriptive and visualization capabilities along with the availability of userfriendly software have enabled plant scientists to examine any two-way data by a click on a computer button. Despite widespread use, the validity and limitations of biplot analysis have not been completely examined. Here we identify and briefl y discuss six key issues surrounding overutilization or abuse of biplot analysis. We question (i) whether the retention of the fi rst two multiplicative terms in the biplot analyses is adequate; (ii) whether the biplot can be more than a simple descriptive technique; (iii) how realistic a “which-won-where” pattern is identifi ed from a biplot; (iv) what if genotypes and/ or environments are random effects; (v) how relevant biplot analysis is to the understanding of the nature and causes of interaction; and (vi) how much the biplot analysis can contribute to detection of crossover interaction. We stress the need for use of confi dence regions for individual genotype and environment scores in biplots to make critical decisions on genotype selection or cultivar recommendation based on a statistical test. We conclude that the biplot analysis is simply a visually descriptive statistical tool and researchers should proceed with caution if using biplot analysis beyond this simple function. R.-C. Yang, Agriculture Research Division, Alberta Agriculture and Rural Development, #307 , 700-113 St., Edmonton, AB, T6H 5T6, Canada; and Dep. of Agricultural, Food and Nutritional Science, Univ. of Alberta, Edmonton, AB, T6G 2P5, Canada; J. Crossa and J. Burgueño, Biometrics and Statistics Unit, Crop Informatics Lab. (CRIL), International Maize and Wheat Improvement Center (CIMMYT), Apdo. Postal 6-641, Mexico, D.F., Mexico; P.L. Cornelius, Dep. of Plant and Soil Sciences and Dep. of Statistics, Univ. of Kentucky, Lexington, KY 40546-03121. Received 19 Nov. 2008. *Corresponding author ([email protected]). Abbreviations: AMMI, additive main eff ects and multiplicative interaction; ANOVA, analysis of variance; BLUP, best linear unbiased prediction; CI, confi dence interval; COI, crossover interaction; FA(2), factor analytic model with the fi rst two latent factors; GE, genotype × environment interaction; GGE, genotype main eff ects and genotype × environment interaction; GL, genotype × location; GLBM, general linear-bilinear model; MET, multi-environment trials; PC, principal component; PCA, principal components analysis; SHMM, shifted multiplicative model; SREG, site regression model; SVD, singular value decomposition. Published in Crop Sci. 49:1564–1576 (2009). doi: 10.2135/cropsci2008.11.0665 © Crop Science Society of America 677 S. Segoe Rd., Madison, WI 53711 USA All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.