Previous Next

Detecting Multicollinearity

1.     Examine the simple correlation among regressors (independent variables).  The general rule is: if two regressors show a correlation coefficient above 0.7, then multicollinearity may be a problem.

2.     Compare the simple correlation coefficients among regressors to the R2.  If the simple correlation exceeds the R2 then multicollinearity may be a significant problem. 

3.     Compute Variance Inflation Factors (VIFs).  Variables with VIFs in excess of 10.0 are considered problematic.  It may be advisable to eliminate one of the two variables from the analysis.


Previous Next