Question: Assume that there are five available variables, and the best subset of size 3, in terms of Cp, is a proper subset of the best subset of size 4. Determine a lower bound on the value of the F-statistic (with σ2 estimated using all of the variables) for testing that the coefficient of the added (fourth) regressor is zero if C3 - C4 > 1 and both Cp values are less than p (i.e., the Cp values suggest that the four-variable subset is better than the three-variable subset). While recognizing that n and the significance level are not given, does your lower bound on the F-value suggest that a subset can be selected from the amount by which Cp is less than p? Or should the relationship between Cp and p simply be used as a rough guideline?