models depend on x in a different way. But no direct information on the magnitude of the difference between the mean
responses of the two models is provided by this approach of hypotheses testing, whether H0 is rejected or not.
Our approach is to assess directly the difference between the mean responses of the two models