Power of mental health nursing research: A statistical analysis of studies in the International Journal of Mental Health Nursing

Having sufficient power to detect effect sizes of an expected magnitude is a core consideration when designing studies in which inferential statistics will be used. The main aim of this study was to investigate the statistical power in studies published in the International Journal of Mental Health Nursing. From volumes 19 (2010) and 20 (2011) of the journal, studies were analysed for their power to detect small, medium, and large effect sizes, according to Cohen's guidelines. The power of the 23 studies included in this review to detect small, medium, and large effects was 0.34, 0.79, and 0.94, respectively. In 90% of papers, no adjustments for experiment-wise error were reported. With a median of nine inferential tests per paper, the mean experiment-wise error rate was 0.51. A priori power analyses were only reported in 17% of studies. Although effect sizes for correlations and regressions were routinely reported, effect sizes for other tests (χ2-tests, t-tests, ANOVA/MANOVA) were largely absent from the papers. All types of effect sizes were infrequently interpreted. Researchers are strongly encouraged to conduct power analyses when designing studies, and to avoid scattergun approaches to data analysis (i.e. undertaking large numbers of tests in the hope of finding ‘significant’ results). Because reviewing effect sizes is essential for determining the clinical significance of study findings, researchers would better serve the field of mental health nursing if they reported and interpreted effect sizes