EcoSta 2019: Start Registration
View Submission - EcoSta2019
Title: Linear discriminant analysis with high dimensional mixed variables Authors:  Binyan Jiang - The Hong Kong Polytechnic University Shenzhen Research Institute (Hong Kong) [presenting]
Abstract: Discriminant analysis is considered with both high dimensional discrete and continuous variables. Under the location-scale Gaussian model, we show that the optimal classification direction relies on the continuous variables via a functional coefficients and the contribution of the discrete variables appears in the intercept. Direct estimation methods is then proposed to estimate the directions and the intercepts. Asymptotic results on the estimation accuracy and the misclassification rates are established and preliminary numerical results will be presented to illustrate the competitive performance of our approach.