This blog explains the concept of power in research and discusses the importance of ensuring that an experiment has sufficient power. It explains that a study requires sufficient power to be able to be confident that you can detect a difference between the groups you are comparing, if a difference truly exists.
The blog explains that there are two main reasons why a study may not show a significant difference between the groups being studied:
There really was no significant difference (a true negative result)
There was a difference but the study failed to detect it because the study lacked power.
The three important factors that influence power are then outlined: 1) sample size: in that it is hard to detect differences in a small sample 2) anticipated effect size: bigger effects are easier to detect than small effects 3) the significance level of the statistical test one uses: you are more likely to find a significant effect if you set a more lenient minimum significance level (e.g. 10% rather than 5%).
The blog then explains how one can choose these parameters. That is, if some of the variables are known (e.g. anticipated effect size and the level of significance you will use), you can then use a power calculator to determine an adequate sample size.
Finally, the blog explains that while it is important that studies are adequately powered (primarily by having a sufficiently large sample size), you can have ‘too much’ power. In that, if you increase your sample size to infinity (hypothetically speaking), then almost any difference between two groups will be found to be significant, no matter how small or irrelevant. Read the blog
Students 4 Best Evidence (S4BE) is a growing network of students from around the world, from school age to university, who are interested in learning more about evidence-based healthcare (EBH). The network is supported by the UK Cochrane Centre. In addition to the website, the S4BE has a Facebook group and Twitter feed. For more information, read Selena Ryan-Vigs blog which introduces Students 4 Best Evidence.
