Advances and Applications in Statistics
Volume 40, Issue 2, Pages 133 - 156
(June 2014)
|
|
COMBINING OF DIMENSION REDUCTION REGRESSION METHODS
Magda M. M. Haggag
|
Abstract: In this paper, combining dimension reduction methods is proposed for regression models. The proposed combining method is called a smaller dimension reduction direction (SDRD). Some methods of dimension reduction are used and combined by a proposed weighting function. The proposed weighting function is based on the inverse variability of the central subspace of each estimation method. The numerical analysis of real-life data is used and the best results are obtained in the form of less average squared distance between the two directions for a one component model. Also, numerical analysis of simulated data gave best results of the proposed combined estimator in the form of more information and less squared risk. The results are shown for different disturbance factors. It is found that our proposed combined estimator combated the individual methods for large values of the disturbance factor. |
Keywords and phrases: central subspace, combining regression models, dimension reduction, directions, weighting function. |
|
Number of Downloads: 344 | Number of Views: 1089 |
|