My SciELO
Services on Demand
Journal
Article
Indicators
- Cited by SciELO
Related links
- Similars in SciELO
Share
Revista Ciencias Técnicas Agropecuarias
On-line version ISSN 2071-0054
Abstract
DEL VALLE MORENO, Juan and GUERRA BUSTILLO, Walkiria. The Multicollinearity in Multiple Lineal Regression Models. Rev Cie Téc Agr [online]. 2012, vol.21, n.4, pp.80-83. ISSN 2071-0054.
This work is about the Multicollinearity problem between the regressive variables in a Multiple Lineal Regression Model. There are many ambiences in the Agro sciences in which this problem can be found. So, in this work and to create, in the data which is really necessary for this study, a multicollinearity situation, explicative variables were generated in a way that at least two of them had a certain degree of dependency, that is to say ¨an almost lineal combination. Once these conditions are created, an analysis of the Multicollinearity is established taking into consideration the symptom, the diagnose and the treatment.For the symptom, the correlation between regressive variable pairs, the F partial and total test on regressive coefficients, the standard error in each estimator and the determination of the coefficient, among others, were deeply analyzed.For the diagnose, a diagonalization of the correlation matrix was used as well the study of the last own values that gives us a precise information.For the treatment, the Ridge Regression and the Regression of Principal Components were used which resulted very efficient to accurately describe the estimators in the Multiple Lineal Regression Model.
Keywords : multicollinearity; multiple lineal regression model; ridge regression; regression of principal components.