VIF (variance inflation factor)

how strongly an X column overlaps with the others

VIF_j = 1 / (1 − R²_j)

In multiple regression, R²_j is the R² of predicting X column j from all the other X columns. A VIF of 1 means no overlap; above about 5 the separate effects of correlated columns cannot be estimated reliably.

Designed experiments avoid the problem because their factor columns are orthogonal.

Calculate it

Regression and Correlation: Simple and multiple regression, correlation, R², p-values and residual plot.

Related terms

• R² (coefficient of determination): share of the variation the model explains