Elastic net regularization is a popular solution to the overfitting problem, where a model fits training data well but does not generalize easily to new test data. Depending on the particular parameters chosen for the elastic net model, some or all of the regressors are preserved, and their magnitudes are reduced.

Below, section describe EViews tools for estimation of elastic net regression, and the special cases of ridge and Lasso models. You may perform estimation over a single lambda penalization parameter and a grid search over multiple penalization parameters. When multiple parameters are used, EViews also supports options for automatic generation of penalization parameters, as well as cross-validation tools for choosing the parameter with the lowest error.

Following estimation, EViews offers specialized views of tables of the coefficients and other summary statistics, graphs of coefficient evolution with respect to the penalty parameter and other statistics, and diagnostics for cross validation.

The discussion includes brief background for elastic net, Lasso, and ridge regression methods. More detailed discussions may be found in Zou and Hastie (2005) and the textbook by Hastie, Tibshirani, and Friedman (2001).