Variance expansion factor (VIF) is a measure of the severity of complex (multiple) * * * linearity in multivariate linear regression model. It represents the ratio of the variance of the estimator of regression coefficient to the variance when the independent variables are assumed to be non-linearly correlated. Assuming that the model has been centrally standardized, the covariance matrix of the regression coefficient estimator is, where the error term of the centrally standardized model and the variance of the independent variable correlation matrix, so the variance of the regression coefficient estimator of the centrally standardized model is equal to the product of the variance of the error term and the k-th diagonal element in the matrix. This second factor is called variance expansion coefficient, which is called VIFk. It can be proved to be the decisive coefficient between the k-th independent variable and the rest independent variables. Therefore, the higher the correlation between the k-th independent variable and other independent variables, that is, the closer to 1, the greater the corresponding VIFk. On the other hand, if the correlation with other independent variables is low, VIFk is close to 1.