Statistical Guideline No. 5. Include Results of a Power Analysis; if a Power Analysis Was Not Performed, Describe the St
- PDF / 159,100 Bytes
- 2 Pages / 595.276 x 790.866 pts Page_size
- 108 Downloads / 151 Views
INTEGRATIVE REVIEW
Statistical Guideline No. 5. Include Results of a Power Analysis; if a Power Analysis Was Not Performed, Describe the Stopping Rule for Recruitment Suzanne C. Segerstrom 1
# International Society of Behavioral Medicine 2020
Abstract This is one in a series of statistical guidelines designed to highlight common statistical considerations in behavioral medicine research. The goal is to briefly discuss appropriate ways to analyze and present data in the International Journal of Behavioral Medicine (IJBM). Collectively the series will culminate in a set of basic statistical guidelines to be adopted by IJBM and integrated into the journal’s official Instructions for Authors, and also to serve as an independent resource. If you have ideas for a future topic, please email the Statistical Editor, Suzanne Segerstrom at [email protected]. Keywords power . effect size . sample size . stopping rule
The Statistics Guru Cohen [1] was disappointed in that he expected that “methods sections in research articles in psychological journals would invariably include power analyses . . . . Indeed, they almost invariably do not” (p. 155). Decisions about the smallest effect size of interest and the sample size—critical elements determining a study’s power—should be transparent. Therefore, the fifth statistical guideline dictates that all empirical articles should report on those decisions and include details of their power analysis. Power is the probability that the collected data will result in rejection of the null hypothesis and further reflects the probability of a type II error (β), because power is equal to 1-β. Power is itself a function of the probability of a type I error (α), sample size, and the effect size specified (e.g., the magnitude of a correlation). Any time the effect size is not 0 and the null hypothesis is not rejected, a type II error has occurred. However, we typically do not care about an effect size of, say, r = 0.0001, and therefore we do not care if we make a type II error. How we achieve adequate power, therefore, will depend on the smallest effect size we do care about. Typically, α is set * Suzanne C. Segerstrom [email protected] 1
Department of Psychology, University of Kentucky, 125 Kastle Hall, Lexington, KY 40506-0044, USA
at 0.05 (two-tailed), and the smallest effect size we care about is determined either theoretically or empirically. Therefore, the last remaining element that controls power is sample size. Insufficient power has plagued psychology and related fields for a long time. In 1960, analyses of power in the designs of studies in Journal of Social and Clinical Psychology yielded an average of 0.48 (where a conventional target is 0.80). In 1984, studies in the Journal of Abnormal Psychology yielded an average of 0.37 [1]. Cohen [1] hoped that by providing “a short rule-of-thumb treatment of necessary sample size [he] might make a difference” (p. 156). However, his hopes were misplaced. A 2017 report found that power in psychology studies ranged from a mean of 0.23 for
Data Loading...
