A0856
Title: A new lasso refitting strategy
Authors: Liu Guo - Waseda University (Japan) [presenting]
Abstract: The least absolute shrinkage and selection operator (lasso) is a popular method for high-dimensional linear regression under sparsity. However, it has limitations such as bias and non-negligible prediction error. To address such disadvantages, many refitting strategies have been proposed and studied. A novel Lasso refitting method is introduced. Intuitions for this method are provided, proving mathematical properties and conducting numerical studies to demonstrate its advantages. The results show that the new estimator outperforms the standard Lasso method, both theoretically and empirically, which suggests its potential as a powerful tool in high-dimensional statistical modeling.
