The nonlinear AIDS model was estimated using iterative seemingly unrelated regressions with homogeneity and symmetry imposed. Model goodness-of-fit is validated by average R2 values of 0.71 indicating good explanatory power of the AIDS model. Durbin-Watson statistics indicate serial correlation is present in four out of seven equations.
Test results for weak separability at sample means using the AIDS model disaggre- gated by fat content are presented in Table 5. We are aware of the potential drawbacks of the Wald test to test for weak separability, but chose to apply its adjusted version to avoidlarge numbers of additional nonlinear restrictions under which the AIDS model failed to converge (Moschini et al 1994)