When multiple variables are associated with a response, the interpretation of a prediction equation is seldom simple.

First, multiple linear regression models are considered and the design matrices are allowed to be different. Second, the predictor variables are either unconstrained or constrained to finite intervals ...