Improved Ridge Regression Estimators for Binary Choice Models: An Empirical Study
Pages 257-265
Kristofer Månsson, B.M. Golam Kibria and Ghazi Shukur
DOI: http://dx.doi.org/10.6000/1929-6029.2014.03.03.5
Published: 05 August 2014

Abstract: This paper suggests some new estimators of the ridge parameter for binary choice models that may be applied in the presence of a multicollinearity problem. These new ridge parameters are functions of other estimators of the ridge parameter that have shown to work well in the previous research. Using a simulation study we investigate the mean square error (MSE) properties of these new ridge parameters and compare them with the best performing estimators from the previous research. The results indicate that we may improve the MSE properties of the ridge regression estimator by applying the proposed estimators in this paper, especially when there is a high multicollinearity between the explanatory variables and when many explanatory variables are included in the regression model. The benefit of this paper is then shown by a health related data where the effect of some risk factors on the probability of receiving diabetes is investigated.