I'm about to use variance partitioning to interpret my results of a given linear regression model and across models and have come across various criticisms of it most notably by Pedhazur*. Also, the criticisms are of both the approaches to VP - commonality analysis and incremental partitioning of variance. From what I understand the former is not as "dangerous" as the latter.
In short, the criticisms are related to the high dependence on R^2 and multicollinearity. Furthermore, the criticism is about how meaningful it is to attribute explanatory power to individual variables in a multivariate situation.
Since I will be using this model for the first time, and even after reading other posts on this and other forums/literature, I'm having trouble wrapping my head around the criticisms. I'd really appreciate any clear explanation of these or other criticisms.
Thanks in advance for your help!
*Multiple regression in behavioral research : explanation and prediction (1982), Holt, Rinehart, and Winston.
In short, the criticisms are related to the high dependence on R^2 and multicollinearity. Furthermore, the criticism is about how meaningful it is to attribute explanatory power to individual variables in a multivariate situation.
Since I will be using this model for the first time, and even after reading other posts on this and other forums/literature, I'm having trouble wrapping my head around the criticisms. I'd really appreciate any clear explanation of these or other criticisms.
Thanks in advance for your help!
*Multiple regression in behavioral research : explanation and prediction (1982), Holt, Rinehart, and Winston.