Multicollinearity

202211201926
Status:
Tags: Statistics Mathematics

• Phenomenon where a predictor variables can be linearly predicted from others with substantial degree of accuracy.

Effects

• Coefficient estimates change erratically in response to small changes of model or data
• Doesn’t reduce predictive power or reliability of model
  • Only affects calculations regarding individual predictors

Detection

1. Large changes in estimated regression coefficients when a predictor is added or deleted
   • Rule of thumb:
2. Insignificant regression coefficients for affected variables in multiple regression, but significant F-test

Collinearity Diagnostics in SPSS

1. Look at VIF in Coefficients table
   • If <10, no multicollinearity

---

References