Looking for Power: The Difficulties and Possibilities of Finding Participants for Braille Research

Abstract

Studies that are designed to meet goldstandard criteria or are scientifically valid are more likely to get funded (Albro, 2008). They also can be used as evidence for preferred practice, which is required by the No Child Left Behind Act (2002). In addition, experimental or gold-standard research is the most powerful tool that researchers have for establishing cause-and-effect relationships (Gall, Gall, & Borg, 2007). The National Professional Development Center on Autism Spectrum Disorders (2008, n.p.) defines research that meets the standards for evidence-based practice as follows: * randomized or quasi-experimental design studies. Two high quality experimental or quasi-experimental group design studies, * single-subject design studies. Three different investigators or research groups must have conducted five high quality single subject design studies, or * combination of evidence. One high quality randomized or quasi-experimental group design study and three high quality single subject design studies conducted by at least three different investigators or research groups (across the group and single subject design studies). PLANNING FOR POWER Power is the complement to beta (power = 1-beta) and is related to the ability to avoid a Type II error. In other words, power analysis is conducted to avoid determining that an intervention being studied was not successful when, in fact, it was successful, but the study was just not sensitive enough to detect the success. Greater power reduces the chances of mistakenly deciding that the intervention was unsuccessful. Power analysis, in its most basic form, is a calculation of the different design characteristics of a study to determine if certain aspects of the study are robust enough to detect a significant difference between groups if a significant difference exists. These characteristics include effect size, the difference between the means of the groups being compared relative to the variability within those groups; alpha, the level of risk that statistically significant differences are found when they do not actually exist; and the number of participants in the study, which affects how accurately the sample represents the population as a whole (Lipsey, 1990). Power can be calculated using a formula, but is easily determined by reviewing power tables that are found in statistical reference books or software online. By considering and calculating power and including design features that increase power when planning a study, the chance of failing to find significant differences between the groups when significant differences exist, simply because the research design is not stringent enough, is diminished. Calculating power, which can be used when planning studies, can also be applied post hoc to studies that have already been completed (Cohen, 1988). By looking at the number of participants and the effect size and then referring to a power table, one finds that most studies are underpowered. In a review of 85 articles published in a top educational research journal, Brewer (1972, cited in Lipsey, 1990) found that the average power for a medium effect size in educational research was .47 when the desired power should be in the range of .80 or higher. Knowing that most studies are severely underpowered, it is safe to say that many studies may find that successful interventions are unsuccessful. Ferrell, Mason, Young, and Cooney's (2006) meta-analysis of research on braille, which reviewed 652 studies and included only the 20 studies that met the criteria for analysis, created a framework for looking at the power of studies in the area of braille, since it provided the number of participants and the effect size for each study. Table 1 presents examples of the power for at least one variable of each of 14 studies that were analyzed as part of the meta-analysis for which power was calculable. The power range of these examples is greater than . …