The Coefficient of Determination in the Ridge Regression
- Rodríguez Sánchez, A.
- García García, C.
- Salmerón Gómez, R.
- García García, C.B.
ISSN: 2171-892X
Year of publication: 2018
Issue: 26
Type: Article
More publications in: Anales de ASEPUMA
Abstract
The coeffcient of determination, denoted R2, represents the proportion of total variation of dependent variable explained by the model. Thus, it is a relevant measure in linear regression that shows the predictive capacity of the estimated model. When the partition of the sum of squares (SS) holds in ordinary least squares (OLS) estimation, the R2 can be expressed as the ratio of the explained variance to the total variance. However, when there is a problem of collinearity, the OLS estimation presents unstable results and others alternative estimation methods are applied such as the ridge regression. This work reviews if the partition of the sum of squares (SS) is veri ed in the ridge estimation and establishes how the coecient of determination should be de ned in this estimation method.
Bibliographic References
- Hoerl, A. and R. Kennard (1970). Ridge regression: Biased estimation for nonorthogonal problems. Technometrics 12 (1), 55–67.
- McDonald, G. (2010). Tracing ridge regression coefficients. Wiley Interdisciplinary Reviews: Computational Statistics 2 (6), 695–703.
- Stock, J. and M. Watson (2012). Introducci´on a la Econometr´ıa (3a Edici´on). Madrid: Ed. Pearson.
- Uriel, E., A. Perir´o, D. Contreras, and M. Molt´o (1997). Econometr´ıa: El Modelo Lineal. Madrid: Ed. Alfa Centauro.
- Wooldridge, J. (2009). Introductory Econometrics: A modern approach. Canada: South-Western Cengage Learning.