Coping With Collinearity

Authors :

Collinearity is very common in linear regression. The common methods for diagnosing the disturbance , such as evaluating parameter instability when variables are removed from the specification are only suggestive. Recent developments are reviewed which assist in diagnosing collinear disturbances. These include condition indexes and variance proportions decompositions and are available in a number of statistical packages. Some corrective strategies are also examines. In general, it is not correct to simply drop variables from the specification unless they are redundant to the program logic and underlying theory.

Log in to post comments