Chapter 12.2, Problem 6E

Refer to Exercise 12.5. Consider two methods for selecting the dosages. Method 1 assigns three rats to the dosage x = 2 and three rats to x = 5. Method 2 equally spaces the dosages between x = 2 and x = 5 (x = 2, 2.6, 3.2, 3.8, 4.4, and 5.0). Suppose that σ is known and that the relationship between E(Y) and x is truly linear (see Chapter 11). If we use the data from both methods to construct confidence intervals for the slope β₁, which method will yield the longer interval? How much longer is the longer interval? If we use method 2, approximately how many observations will be required to obtain an interval the same length as that obtained by the optimal assignment of method 1?

12.5 Suppose that we wish to study the effect of the stimulant digitalis on the blood pressure Y of rats over a dosage range of x = 2 to x = 5 units. The response is expected to be linear over the region; that is, Y = β₀ + β₁x + ε. Six rats are available for the experiment, and each rat can receive only one dose. What dosages of digitalis should be employed in the experiment, and how many rats should be run at each dosage to maximize the quantity of information in the experiment relative to the slope β₁?