This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: 3.9 In this study, the response of interest is the blood pressure two hours after the application of a treatment minus the blood pressure just before the treatment is applied, i.e. the reduction in blood pressure after two hours. We want to assess the effects of Drug A on blood pressure against the baseline of placebo effect. A placebo effect is a therapeutic effect due to a patient’s belief that a treatment will change his or her condition (Moore and Notz, 2005; Piantadosi, 2005; Meinert, 1986). Thus, we have two treatments: Drug A and placebo. For patients receiving the treatment called placebo, an inert pill is given. Aside from the treatment used, the characteristics of a patient can significantly affect changes in the blood pressure two hours after it was measured. Thus, the two treatments should be compared in blocks, and this leads to a paired comparison design. Each block should consist of two patients who are most similar to each other in all importance baseline characteristics (Meinert, 1986) among all patients that can be chosen for the experiment. One of the important characteristics is the mean level of the patient’s blood pressure. Clearly, we would expect the drug to have less effect on the blood pressure of patients with blood pressure only slightly higher than normal than on the blood pressure of patients with blood pressure much higher than normal. The number of blocks is essentially a sample size problem and so, we choose the number of blocks based on the power of the paired -test, the statistical test of no difference between treatments. Our power must be adequate for detecting an effect in Drug A that is practically significant. Our experiment design is flawed without randomization. We randomly assign one patient in each pair to receive Drug A and the other patient in each pair to receive a placebo. Assuming that only one physician is available to measure the blood pressure of patients, we would also randomize the order in which the pairs of patients are treated. In carrying out the experiment, no patient should be informed about the treatment being assigned to them to avoid biasing the effect of the placebo treatment downwards (Moore and Notz, 2005). If patients know that they are being given a placebo, they would certainly expect no improvement in their condition. It is also possible that a physician involved in the experiment might have preconceived notions about Drug A. As such, the physician should not be told about the specific treatment that is assigned to each patient. Clinical experiments such as this where neither patient nor physician knows the treatments assigned to the patient are called double-blind experiments (Moore and Notz, 2005). Errors incurred in the measurement process must also be controlled. Thus, the physician should use a properly calibrated equipment to measure blood pressure. In addition, it is important to measure the blood pressure of each pair of patients in random order after two hours. It is also important to control the activity of each patient in the two order after two hours....
View Full Document
- Spring '08
- Regression Analysis, main eﬀects