Multicollinearity
Notes and Ideas:
Definition
- The problem where the X matrix is unstable
- The Variance of
are huge - Do not efffect prediction
- It would effect the model itself
How to Deal with it:
- Using of centered data for predictor varaibles when using polynomial regression and other regressions with interactions
- Dropping one or more predictores
- Add some cases that may break the pattern of multicollinearity
- Use different data sets to estimate different
vs outercorrelation