USING A PROPOSED METHOD FOR WAVELET SHRINKAGE TO ESTIMATE THE TUNING PARAMETER OF PENALIZED LINEAR REGRESSION
The tuning parameter selection strategy for penalized estimation is crucial to identifying a model that is both interpretable and predictive. However, popular strategies (e.g., wavelet shrinkage is proposed for effectively handle in these issues) tend to select models with more predictors than necessary. This paper proposes a simple estimate for tuning parameters based on wavelet shrinkage of penalized method (Ridge and Elastic-Net) compared with the classic penalized method depending on the tail probability behavior of the response variables and using simulation experiments for (10%) data with contamination and real data. The comparing results between the proposed method with a classic penalized method based on the statistical criterion (MAE and MSE). It was concluded that the wavelet shrinkage of penalized method gives the best results and a more accurate classical method for all simulations and real data based on (MAE and MSE) criteria.