The comparison of stability parameters according to the Finlay-Wilkinson, Eberhart-Russell and AMMI model

Abstract

It is generally known that a phenotype is formed on the basis of the capacity of its genotype affected by environmental factors. Yield stability is statistically expressed through the GxE interaction. The interaction is a part of the trial variation that includes the factor of the genotype and the factor of the environment. The presence of the qualitative interaction significantly makes difficult the selection of the best genotypes in the process of breeding. As long ago as the early 1960s, many researchers noticed that the application of linear regression could, in many cases, define adaptability and stability of a genotype. However, in cases when data are not well adjusted to the linear model, better results can be gained by the application of the multivariate models (PCA, AMMI, GGE). The estimated maize hybrid stability was compared in this study by linear (Finlay and Wilkinson and Eberhart and Russell models) and multivariate statistical models (AMMI model). Very similar results were obtained by all three models. It was probably due to one dominant factor in the hybrid x environment interaction (different precipitation sums for the environments and the length of the growing season for hybrids). The advantage was given to the AMMI model, as the model with one axis analyses the interaction equally well as linear models, while the model with two axes extracted another part of systemic variation of the interaction effect not adjusted to the linear model.

Opšte je poznato da se fenotip formira na osnovu kapaciteta svog genotipa pod uticajem faktora spoljašnje sredine. Prisustvo kvalitativne interakcije značajno otežava izbor najboljih genotipova u procesu selekcije. Još su ranih šezdesetih godina istraživači primetili da se primenom linearne regresije može, u velikom broju slučajeva, definisati adaptabilnost i stabilnost genotipa. Međutim, u slučajevima kada podaci nisu dobro prilagođeni linearnom modelu bolji rezultati se postižu primenom multivarijacionih modela (PCA, AMMI, GGE). U ovim istraživanjima izvršeno je poređenje procene stabilnosti hibrida kukuruza pomoću linearnih (Finlay i Wilkinson, Eberhart i Russell, modeli) i multivarijacionog modela statistike (AMMI model). Sva tri modela su dala veoma slične rezultate. Prednost se ipak daje AMMI modelu jer model sa jednom osom podjednako dobro kao i linearni modeli analizira interakciju, dok je model sa dve ose izdvojio još jedan deo sistematske varijacije interakcijskog efekta koji nije prilagođen linearnom modelu.