The mean and the covariance matrix are parameters considered by the supervised classification method Lineal Discriminant Analysis, also known as LDA. This study explores a robust and nonparametric approach to estimate them. The main contribution is the estimation of the variability structure as the product between a non-parametric estimation of the correlation matrix and a robust estimator of the standard deviations. The performance of the proposed method was evaluated with synthetic data, simulated predictors and previously known labels while incorporating outlier-like datapoints. The novel LDA version introduced, is compared against the classic and in every scenario presented a better performance in terms of false positive rate, accuracy, sensitivity and ROC curve. Results of said research were published in Colombia's National University's XII Coloquium of Statistics The Poster down below sumarises the main results of the study and was presented alongside with the article during the event.