Muhammad F., Ghafoor A.
Abstract
In classification of traits, it is observed that the plant traits such as pods per plant, pod length and biological yield per plant have positive contribution towards mash grain yield. Three regression procedures i.e. best subset regression, principal component regression and ridge regression were tried as in the first phase and for pod length, approximately same positive quantitative effect was observed towards mash grain yield for all three predictions models. Remaining mash plant traits contribute negatively towards mash grain yield. Also on the basis of criterion of goodness of fit, it was observed that best subset regression model is best and stable as compared to principal component and ridge regression prediction models, both on the basis of original data and simulation procedure.
A simulation procedure adopted to test the reliability of the results by generating random samples from normal (0,1) exponential (1) and uniform (0,1) distributions revealed that estimated effect for pod length for uniform (0,1) distribution tends to very close to original results as compared to other two distributions.
Key words:
plant traits, yield, principal component regression, Bayesian regression, simulations.