I have to evaluate a Type II second-order construct (reflective-formative). Following the common two-step approach, I already evaluated the measurement (outer) model. Proceeding with the structural (inner) model, I want to check for multicollinearity between the first-order dimension. As this is typically done using the variance inflation factor (VIF), I wonder how to determine these within SPSS/AMOS.
I have read of two different approaches so far:
1) The VIFs correspond to the diagonal of the inverted correlation matrix between the first-order dimensions.
2) VIF = 1/(1-R^2) within linear regression
Both approaches lead to different values, so I wonder which approach is correct.
Especially since for the second approach I first have to create a new variable for each first-order dimension from its indicators, whereas the correlation matrix for the first approach is directly available in the AMOS output.