1. The stepwise model (Table 6) includes the explanatory variable Sq. Temp Deviation, the square of the difference between room temperature and 75 degrees. Why 75? Add Room Temp itself to the regression, and then use the resulting estimates to find the optimal center for this quadratic effect.
2. Suppose you initialize backward stepwise regression from the saturated model. Use Plant rather than Manager in this model. Must the backward search eventually reach the model chosen by the forward search? Try it and see.