Abstract
The stepwise procedure of selecting variables is common methods reduce the number of variables in linear discriminant analysis. The discriminatory power of variables is measured by adopting Wilk's lambda, which is significantly affected by the presence of outliers. Therefore, a novel robust deterministic minimum covariance determinant (DetMCD) algorithm was applied in this study to increase the resistance of Wilk's lambda against outliers. The lambda was constructed by using the DetMCD estimator. The values of Wilk's lambda were compared through robust and classical methods. The robust Wilk's lambda was applied in a simulation. Results showed the superiority of the novel DetMCD algorithm.