Variable Selection for Discrimination between Low and High Yielding Populations of Indian Mustard

Abstract

Variable Selection is an important problem in classification and discriminant analysis. The selection of important variables for the purpose of discrimination between populations is important from the point of view of time and resources required for the experimentation. Keeping this in view, the present study has been designed to find important characters of Indian mustard which can discriminate between high and low yielding genotypes. Secondary data set on 310 genotypes of Indian mustard recorded for 12 characters was used for discrimination between populations of low and high yielding genotypes of Indian mustard. Three variable selection methods (Univariate t-test, Rao ́s F test for additional Information and Random Forests Algorithm) for classification and discrimination were used and compared. Performance of the methods was assessed in terms of leave one out cross-validation error and out of bag error rate for classification. The Four most important variables for discrimination among genotypes based on seed yield per plants were secondary branches, primary branches, days to maturity and siliqua number on main shoot.