Advances and Applications in Statistics
Volume 48, Issue 6, Pages 391 - 409
(June 2016) http://dx.doi.org/10.17654/AS048060391 |
|
MATHEMATICAL PROGRAMMING MODEL FOR SELECTING GENERALIZED RIDGE REGRESSION PARAMETER
Aya Farag, Ragaa Kassem and Ramadan Hamed
|
Abstract: Ridge regression technique is used to estimate the regression coefficients instead of least squares method when there is multicollinearity in the data. Generalized ridge regression is more realistic than ordinary ridge regression as it assumes separate ridge parameters for each explanatory variable according to its nature. This study proposes nonlinear mathematical programming model in order to estimate the generalized ridge parameter such that to minimize both generalized cross validation and mean square error functions. The results of the proposed mathematical programming model are checked through a simulation study. In addition, the suggested model is compared to other models. The proposed model has good performance in terms of measures, mean square error and generalized cross validation. |
Keywords and phrases: generalized ridge regression, mean square error, generalized cross validation function, mathematical programming model. |
|
Number of Downloads: 422 | Number of Views: 1374 |
|