Question: It was mentioned that the combined Box-Cox and Box-Tidwell approach might be followed by the use of the modified Box-Cox approach for a specific set of regressor transformations in an effort to see if nonnormality and/or heteroscedasticity that seems to be present in all models identified by the two-stage procedure might be removed. Examine this possibility for the Minitab tree data by using the regressor transformations suggested in Table, searching for a different value of λ in increments of .05 from log (Y). Is an acceptable model apparent? Would the results likely have been different if you had used logarithmic terms in both regressors?
