Draft 5-Experimental Power
In order to determine the power of an experiment, the experiment must be conducted several times. If the experiment is successful or shows similar results the majority of the time, the experiment would have a great power. Initially, before conducting an experiment several or dozens of times, the initial experiment must be well thought out. An experiment should have a high (if cost is an issue, then reasonable) amount of replicants. All experiments will have an experimental noise which will cause variation, but if there is an alarming amount of variation in the initial experiment, there is something wrong with the experimental design, there could be another variable interfering in the experiment which isn’t realized, etc. A simple experimental design has more power than a very complex one. The P-value cutoff should be around 0.05, meaning that the majority of the time, the experiment is accurate and the results are more easily proven. Finally, an experiment which shows a strong treatment effect will have greater power than an experiment which does not.
Recent comments