Abstract
In the context of regression modeling, when the number of predictors are relatively large with respect to sample size, the least absolute shrinkage and selection operator (LASSO), adds some level of sparsity to the estimation of parameter vector to improve the prediction. When the coefficients are subjected to lie in a subspace hypothesis, we propose to use a restricted LASSO estimator along with its preliminary test and shrinkage versions. Asymptotic distributions with properties of the newly defined LASSO-typed estimators are derived and some graphs are depicted to show the performance of asymptotic relative efficiencies.
Type
Publication
10th Seminar on Probability and Stochastic Process
