Correlation, Path-coefficient, Normal and Stepwise Regression Analyses Via Two Cycles of Pedigree Selection in Bread Wheat (Triticum aestivum L).

Correlation, path- coefficient and regression analyses are important statistical tools which can help breeders to characterized the crop populations during the selection program and select the desirable genotypes of high yield. The current study was aimed to assess these analyses during three successive seasons i.e., 2012/2013, 2013/2014 and 2014/2015 at Fac. Agric. Farm, Assiut University, Egypt. The genetic materials were two segregating bread wheat populations i.e. Debeira x Sahel and Giza 165 x Sakha 93. Grain yield/plant revealed positive and high phenotypic correlation with each of number of spikes/plant, biological yield/plant and weight of spikes/ plant in base and the two cycles of pedigree selection for grain yield/plant in both populations, respectively. It is clear that the strong correlation occurred between grain yield/plant and weight of spikes/plant in all cases. The partitioning of phenotypic correlation into direct and indirect effects by path analysis revealed that the highest direct effect on grain yield/plant was exerted by weight of spikes/plant in base and the two cycles of pedigree selection for grain yield in both populations. Simple regression analysis revealed that the weight of spikes/plant (WS/P) (model no. 6) was superior to other traits and its relative contributions (more than 0.920) in grain yield in both populations across all generations. Multiple regression analysis revealed that, in both populations, the highest contribution in grain yield as more than 96% in base and the two cycles of selection were obtained as done from the model no. 9 which included only two traits i.e. WS/P and threshing index (TI). Meanwhile, those two traits were the main predominant elements in both populations. The stepwise regression analysis revealed to three fitted models for each of base and cycle one as well as two superior models for cycle two of selection in population I. Meanwhile, two fitted models were exhibited for each of base and cycle one and four efficient models released from cycle two of selection in population II. The model no. 1 in all cases included only the WS/P in both populations. It is remark results that the relative contribution of WS/P in grain yield/plant was decreased from 0.921, to 0.843 and to 0.782 and from 0.965, to 0.922 and to 0.840 for base, cycle one and cycle two of selection in population I and II, respectively. Moreover, the model no. 2 in all cases, also, included two traits of WS/P and TI in both populations, which increased the relative contribution for grain yield/plant from 0.921 to 0.988 in base, from 0.843 to 0.969 in cycle one and from 0.782 to 0.993 in cycle two of pedigree selection in population I. As well as from 0.965 to 0.994 in base, from 0.922 to 0.989 in cycle one and from 0.840 to 0.997 in cycle two of selection in population II.