Using Power Analysis to Determine the Appropriate Number of Subjects




Sample Paper Excerpt
A reviewer is reading over a proposal of a study investigating the outcomes of two different types of substance abuse treatment for people with Psychoactive Substance Use Disorder (PSUD). The research design is a Randomized Clinical Trial (RCT) where the subjects are randomly assigned to two different treatment conditions. One condition is an innovative type of group therapy; the other group is traditional treatment which will serve as the control condition. The subjects were referred to treatment by the criminal justice system and were long-term addicts strongly habituated to drug use. The new group therapy is designed with consideration given to the unique characteristics of this treatment population. Since their PSUD is severe, the researchers expect a large effect size. The proposal must include power analysis to determine the appropriate number of subjects to include in each group. Given the novel nature of the research, the researchers found no relevant studies published in the literature that would give them any indication of the size of the effect difference they might find (as measured by pooled variance). The researchers were able to find a study from a parallel literature source whose subjects were all working adults whose treatment was a part of an Employee Assistance Program (EAP). This treatment population is unique because their PSUD was not severe enough to cause them to leave the workforce so the researchers powered the study to detect small effect differences. The following power statistics based on the study of people with less severe PSUD-that is, the study of subjects enrolled in an EAP-based program Sample Size of 75 per Group Pooled Variance: 50 60 70 80 90 Power: 96.3 92.8 88.7 84.3 79.8 A reviewer reading the proposal has a concern about the power calculation proposed by the authors of the proposal.

What is the reviewer's concern?
The reviewer feels a power calculation is unnecessary. The reviewer feels that drawing from dissimilar populations could pose an ethical problem. The reviewer feels the power calculation underestimates the size of the sample needed. The reviewer feels the power calculation cannot accurately estimate the size of the sample needed.



Finished?