Add the teaching of power analysis software to your lesson plans. Most classes in advanced statistics focus on software that performs significance tests and model construction. However, power analysis is just as important as these two operations. Instruct students to become familiar with the many programs that allow the calculation of power, such as JMP, SAS, S-PLUS, GPOWER and PASS. If students become familiar with these programs, they will be more likely to use them in the future.
Remind students that performing statistical tests with low power may not yield results. Even when the alternative hypothesis is correct, a significance test may fail to reject the null hypothesis due to low power. If students know to calculate power before performing a statistical hypothesis test, they will know if the test is worth running. This is especially important in educating future researchers who will be planning and performing experiments, the results of which rely on having high power.
Teach students that sample size can be determined by power analysis. If a researcher's education lacked mention of power analysis, it is possible that researcher may be choosing sample sizes for his or her research based on intuition or rules of thumb. In this case, he or she may be unknowingly setting the experiment up for research. Educate students that by deciding on a minimum amount of power for a particular experiment before performing the experiment will allow him or her to calculate an appropriate sample size for the experiment. Teach students to always keep in mind the relation between an experiment's sample size and its power.
Employ the use of examples to show students how a study can fail or succeed due to appropriate power analysis. If educators can give students tangible examples of when power analysis is important in real research, students will be able to more easily make a link between the importance of power analysis and their own research. As an example, show the students how a significance test will almost always fail when power is equal to 10 percent, no matter if the null hypothesis was never correct.