Assessing the importance of an independent variable in multiple regression: is stepwise unwise?

Two procedures for evaluating the importance of an independent variable are reviewed: the allocation of R2 method and the significance method. Multicollinearity severely hampers attempts to evaluate an independent variable using the allocation of R2 method. The significance method relies on assessing the stochastic, "magnitudinal", (or clinical) and practical significance of an independent variable and does not suffer as severely from the ravages of multicollinearity. An example of the limitations of allocations of R2 method in stepwise regression is discussed and another interpretation of the results using the significance approach is suggested.