Robust multicollinearity diagnostic measures based on minimum covariance determinants approach
The classical multicollinearity diagnostic measures are not resistant to high leverage points since their formulation are based on eigen analysis of classical correlation matrix that is very sensitive to the presence of these leverages. The existing robust multicollinearity diagnostics also are not...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | en |
| Published: |
Academy of Economic Studies
2013
|
| Online Access: | http://psasir.upm.edu.my/id/eprint/35288/1/Robust%20multicollinearity%20diagnostic%20measures%20based%20on%20minimum%20covariance%20determinants%20approach.pdf http://psasir.upm.edu.my/id/eprint/35288/ http://www.ecocyb.ase.ro/Articles20134.htm |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The classical multicollinearity diagnostic measures are not resistant to high leverage points since their formulation are based on eigen analysis of classical correlation matrix that is very sensitive to the presence of these leverages. The existing robust multicollinearity diagnostics also are not able to diagnose the variables which are collinear to each other. In this paper, we proposed robust multicollinearity diagnostic measures based on the Minimum Covariance Determination (MCD), which is a highly robust estimator of multivariate location and scatter. The results of numerical example and simulation study confirmed the merit of our new proposed robust multicollinearity diagnostic measures. |
|---|