A1496
Title: Variable selection based on multivariate regression association
Authors: Jia-Han Shih - National Sun Yat-sen University (Taiwan) [presenting]
Yi-Hau Chen - Academia Sinica (Taiwan)
Abstract: A multivariate regression association measure based on spatial signs is applied for variable selection. This measure quantifies the predictability of a multivariate outcome from a multivariate covariate and is therefore particularly suitable for variable selection. Since the measure is estimated nonparametrically, the corresponding variable selection procedure is also nonparametric. Thus, it can capture complicated dependence structure of response on covariate. Simulation studies are conducted to compare the performance of the proposed method with several existing approaches in the literature under various models. The results show that the proposed method performs competitively with existing methods. Finally, a real dataset is analyzed for illustration.
