Abstract
One of the most important traits that plant breeders aim to improve is grain yield which is a highly quantitative trait controlled by various agro-morphological traits. Twelve morphological traits such as Germination Percentage, Days to Spike Emergence, Plant Height, Spike Length, Awn Length, Tillers/Plant, Leaf Angle, Seeds/Spike, Plant Thickness, 1000-Grain Weight, Harvest Index and Days to Maturity have been considered as independent factors. Correlation, regression, and principal component analysis (PCA) are used to identify the different durum wheat traits, which significantly contribute to the yield. The necessary assumptions required for applying regression modeling have been tested and all the assumptions are satisfied by the observed data. The outliers are detected in the observations of fixed traits and Grain Yield. Some observations are detected as outliers but the outlying observations did not show any influence on the regression fit. For selecting a parsimonious regression model for durum wheat, best subset regression, and stepwise regression techniques have been applied. The best subset regression analysis revealed that Germination Percentage, Tillers/Plant, and Seeds/Spike have a marked increasing effect whereas Plant thickness has a negative effect on durum wheat yield. While stepwise regression analysis identified that the traits, Germination Percentage, Tillers/Plant, and Seeds/Spike significantly contribute to increasing the durum wheat yield. The simple correlation coefficient specified the significant positive correlation of Grain Yield with Germination Percentage, Number of Tillers/Plant, Seeds/Spike, and Harvest Index. These results of correlation analysis directed the importance of morphological characters and their significant positive impact on Grain Yield. The results of PCA showed that most variation (70%) among data set can be explained by the first five components. It also identified that Seeds/Spike; 1000-Grain Weight and Harvest Index have a higher influence in contributing to the durum wheat yield. Based on the results it is recommended that these important parameters might be considered and focused in future durum wheat breeding programs to develop high yield varieties.
Highlights
Durum wheat (Triticum durum) is an important crop with an estimated global cultivation area over 13 million hectares that constitute only 5-8% of the world wheat production (Kadkol and Sissons, 2016)
The identification of the minimum, but the most important, parameters by building a parsimonious model to predict yield have significant importance to suggest specific parameters that might be used as selection criteria for future breeding programs to improve crop yield
Germination Percentage (X1), Tillers/Plant (X6), Seeds/ Spike (X8), and Plant Thickness (X9) with R2p = 87.8%, RA2dj = 87.1 and Cp = 3.5 seems to be a good model. These results indicate that above-selected durum wheat traits are the foremost traits for predicting grain yield of durum wheat
Summary
Durum wheat (Triticum durum) is an important crop with an estimated global cultivation area over 13 million hectares that constitute only 5-8% of the world wheat production (Kadkol and Sissons, 2016). Keeping in view the changing food priorities and increasing demand for pasta synthesized products in Pakistan, CIMMYT is making efforts to alter the local market dynamics to promote disease resistance and high yielding durum wheat varieties with improved grain quality (Joshi et al, 2015). Correlation and path coefficient analysis is the important statistical techniques used to assist crop breeding programs to study the direct and indirect impact of yield components on grain yield. Mohammadi et al (2011) applied correlation and regression analysis and identified the existence of high heritability for growth vigor, days to maturity, plant height, peduncle length, number of kernel per spike, flag leaf senescence, spike length, thousand kernel weight and test weight as most effective selection parameters for yield. Through the regression modeling process, the factors Erucic Acid and Pods Length were identified that significantly contribute to increasing the production of the mustard crop in Pakistan (Saleem et al, 2013). Mohsen (2013) used best subset and stepwise regression and identified all the independent variables as significant traits excluding number of seed/ spike and seed weight
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
