استعمال بعض طرائق تقدير الأنموذج شبه المعلمي أحادي المؤشر

Abstract EN

The issue of estimation models and chooses the significance variable is a crucial process in construction of the semiparametric models.

At the beginning of the modeling process usually there is a large number of explanatory variables for the purpose of avoiding the loss of any elements may be important and as a result, the choice of significance variables becomes a necessity not to the fact that the selection process variable model designed to simplify the complex model only but also to improve their ability to predict In this paper we use and compare certain methods that deals with estimating and selecting the significance variables at the same time for the Single Index Model, These methods are (Minimum average variance estimation (MAVE) Method, Minimum average variance estimation (MAVE) Method with LASSO penalty function (LASSO-MAVE) Method, Non-concave penalized least squares method with (SCAD) penalty function method (SCAD – NPLS) method.

Simulation is used to study a small sample behavior for number of models and other cases, based on the comparison criterion (Mean average squared error) we reached that best method is (MAVE-LASSO) for the first simulated model and (MAVE) is the best for the second model.