After creating a linear regression model, quantifying the importance of each variable is often desired, either for academic purposes or to design the most effective interventions. This normally involves decomposing the variance, or, equivalently, R². Using the Shapley Value, known in this literature as the LMV, it is possible to do such a decomposition. This decomposition can be applied to models with interactions and is robust against multicollinearity. It has the interpretation that it is the marginal contribution of each variable to R². This is implemented in macro %DECOMPOSE_R2 using PROCs REG, IML, macro string functions and data step-like macro processing.