In estimating the advertising response function in Example 7.5, we indicated that the sum of squared prediction errors or RMSE could be used as the objective, and we used RMSE. Try using the sum of squared prediction errors instead. Does Solver find the same solution as in the example? Try running Solver several times, each time from a different starting solution in the changing cells, and report what happens. Alternatively, use the Multistart option with the bounds we suggested.