Multivariate statistical analysis of grain yield and agronomic characters in Durum wheat

Correlation, stepwise multiple regression and factor analyses were conducted on grain yield and a number of agronomic characters in the parental, F 1 and F 2 families originating from a 10 X 10 diallel cross in durum wheat. For the F 1 diallel, the correlation analysis indicated that the number of spikes and kernels per plant and 1,000 kernel weight had the highest correlations with grain yield; for the F 2 diallel, the number of spikes and kernels per meter, 1000 kernel weight and plant height showed most striking correlations with same.Stepwise multiple regression analysis indicated that, for the F 1 diallel, number of kernels per plant, 1000 kernel weight and days to maturity were the most potent predictor variables for grain yield, accounting for 78% of its variability. For the F 2 diallel, the number of kernels and number of spikes per meter, 1000 kernel weight and number of kernels per spike were the most potent predictors for grain yield, accounting for 91 % of its variability. Five common factors were extracted which explained 98.8% and 98.1% of the total variance in the F 1 and F 2 diallel, respectively. However, the importance of each of the five factors and the characters which loaded highly on each of them differed from generation to generation.