Transformationen in gemischten Modellen: Anwendung bei der Auswertung einer Versuchsserie

engl. Titel : Transformations in mixed models: application to riskanalysis for a multi-environment trial

Hans-Peter Piepho(1), Charles E McCulloch(2)

(1) Institut fuer Nutzpflanzenkunde, Universitaet Kassel, Witzenhausen, Germany 
(2) Cornell University, College of Agriculture and Life Sciences, Department of Biometrics, Ithaca, NY, USA 
 

An important trait in crop cultivar evaluation is stability of performanceacross environments. There are many different measures stability, most ofwhich are related to variance components of a mixed model. We believe thatcstability measures assessing yield risk are of particular relevance,because they integrate location and scale parameters in a meaningful way.A prerequisite for obtaining valid risk estimates is an appropiate modelfor the distribution of yield across environments. Multienvironment trials(MET) are often analyzed by mixed linear models, assuming thatenvironments are a random sample from a target population, and that randomterms in the model are normally distributed. The normality assumption maynot always be tenable, and consequently, risk estimates may be biased. Inthis paper, we suggest two different approaches to cope with non-normalityin mixed models. The first uses a Box-Cox transformation of the response,while the second is based on nonlinear mixed models. The methods areexemplified using an international wheat yield trial. The importance ofaccounting for non-normality in risk analyses based on MET isemphasized.

Literatur :

Piepho HP 1998 Methods for comparing the yield stability of cropping systems - A review. Journal of Agronomy and Crop Science 180, 193-213.

Piepho HP, McCulloch CE 1999 Transformations in mixed models: application to risk analysis for a multi-environment trial. In preparation

Piepho HP, van Eeuwijk FA 1999 Stability Analysis in Crop Performance Evaluation. In: Kang MS (ed) Crop improvement for the 21st century. Vol. 2. Research Signpost, India. to appear