"A comparative study between LASSO-MAVE method and Adaptive LASSO-MAVE method for variable selection in semi-parametric single index models

Abstract

The semi-parametric single – index model (SSIM) are important tools and basic to treatment the problem of high – dimensional , As it plays an important role in the process of model building and variable selection of significant . in this research has been the use some methods variable selection of automatic modern and that work on estimation vector of parameters β and link function g (X^T β ) With variable selection at the same time for semi-parametric single-index model are LASSO -MAVE method and Adaptive LASSO - MAVE method for aim to improve the accuracy and predict of the model . in order to achieve this aim , it was conducted simulation experiment to show methods preference used in estimation and variable selection for model under study by using different models , different variances , different sample sizes and different correlation values as well as the use of a real data of influencing factors on market value share for the banks sector in the iraqi stock exchange for purpose of comparison and a check from performance these methods in practice . it was reached through simulation experiments and a real data to conclusion showed favorite Adaptive LASSO - MAVE method as it gave better results from LASSO-MAVE method depending on the two criteria Average mean squared error (AMSE) and Average mean absolute error (AMAE) basically for Comparison , and were obtained on results depending on program R- package.