×

Efficient choice of biasing constant for ridge regression. (English) Zbl 1213.90222

Summary: This endeavor developed a mathematical programming model, which has two main objectives minimizing variance inflation factors (VIF) and maximizing coefficient of determination, to yield an efficient biasing constant for ridge regression, which is widely used for multicollinearity problems. The multiobjective structure is handled by means of goal programming. The model allowing analysts preemptive planning is analytical rather than trial and error, which is often the case in determining the biasing constant for ridge regression. The road of approach facilitates decision makers to move in an efficient region for choosing the constant via efficient frontiers stemming from the nonlinear goal programming model. The derived model requiring a few inputs may be embedded in any appropriate statistical software even though it is implemented on MS Excel for a published data set in this study. The results coming from even other many observations are impressive and definitely in accordance with that of the literature. The recipe may be a useful tool for ridge regression analysis for both researchers and industry.

MSC:

90C29 Multi-objective and goal programming

Software:

Excel; AS 223
PDFBibTeX XMLCite