An evaluation of multivariate statistical techniques for the analysis of yield from barley (Hordeum vulgare L.) breeding trials data

Abstract

This project involved two locations (Breda and Tel Hadya) over two seasons (1993 and 1994). Yield was found to have been affected by many factors including environment, genotype and morphological characters. A genotype-environment interaction (GEl) was also discovered. To investigate the influence of morphological characters on yield parameters, multivariate statistical techniques (canonical analysis, factor analysis and multiple regression analysis (linear and exponential)) were used. Multivariate statistical techniques were applied to three hybrids (Hybrid 1, 2 and 3) in replicated field plots at two locations (Breda and Tel Hadya) in two seasons. Canonical analysis and factor analysis revealed a significant relationship between yield parameters and morphological characters. However, this relationship was not significant for each hybrid because there were insufficient data for each hybrid. Stepwise multiple regression analysis showed that plant height, vegetative duration and length of growing season were the significant factors influencing yield parameters, while leafiness was not. The relationship can approximate nonlinear in that it gives more realistic predictions. Consequently, stepwise multiple exponential equation fitted the data better than stepwise multiple linear equation. The relationship between yield parameters and morphological characters was affected by environment but not by genotype.