Abstract:

 Based on the profile least squares method, the LASSO penalty profile least squares estimator is constructed, and the choices of penalty parameter and bandwidth are also discussed. Because the optimization problem is linear, it can be easily implemented. Under some regular conditions, the consistency and asymptotic normality of the estimator for parameter component are investigated. Finally, Monte Carlo simulation studies are conducted to assess the finite sample performance of the proposed variable selection procedures.

Key words: partially linear models; variable selection; asymptotics; LASSO