Unformatted text preview: 1 terms in the cross product so that later when we test that for example the model is just two connected lines at 8, testing the beta =0 for the quadratic term may have an effect on the x 1 linear term. Of course, when there are several ways to write the models, ultimately it depends for our purposes on what we are going to be asked to test and can we test it by testing a subset of betas = 0. If not, we may have to reparameterize! For just the part with A, how about + 1 x 1 + 2 (x 18)x 4 + 3 (x 1 28 2 )x 4 , then for E(y) we would just have our usual way of writing most general model where this Is the model with quantitative variables alone: E(y) = + 1 x 1 + 2 (x 18)x 4 + 3 (x 1 28 2 )x 4 + 4 x 3 + 5 x 1 x 3 + 6 (x 18)x 4 x 3 + 7 (x 1 28 2 )x 4 x 3 where x 3 = 1 if B, 0 otherwise; x 4 = 1 if x1 <= 8, 0 otherwise...
View
Full Document
 Fall '10
 Szatrowski
 x3

Click to edit the document details