2

When calculating the Variance Inflation Factor for a variable $X_j$ in a multiple regression:

$$VIF_j= \frac{1}{1-R^2_{j}}$$

if the variable in question is a dummy variable with 0/1 values, is it appropriate to use a pseudo-$R^2$ for $X_j$, such as McFadden's pseudo-$R^2$, in place of the standard $R^2$ for continuous variables?

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
RobertF
  • 4,380
  • 6
  • 29
  • 46
  • 2
    What you want to use in this situation is the GVIF (Generalized Variance Inflation Factor), as explained here: https://stats.stackexchange.com/questions/294678/multicollinearity-between-two-categorical-variables. – Isabella Ghement May 21 '18 at 00:39

0 Answers0