The Effects of Nonlinearity in Regression Models. Part 2: A Function Describing Nonlinearity.

Abstract : In a previous paper (Draper and Shaw, 1973), the authors provided a way of assessing nonlinearity in nonlinear estimation problems with n observations and p parameters via a function F bar(p,n-p, 1-alpha). This function measured the average deviation from linearity around the usual linearized (1-alpha) F bar(p,n-p, a-alpha) as a function of F(p,n-p,1-alpha), and demonstrate that the coefficients convey, in an intuitively sensible manner, information about the nonlinearity of the model. (Modified author abstract)