Power

David C. Howell

Most of what I currently have on Power comes from software programs that generate samples, compute statistics, and report those statistics. I'll add other kinds of things as I come to them.

• The power of chi-square can be demonstrated in the three-door problem.
• The power of t  tests can be seen with this linked program.
• A particularly interesting example can be seen with the F   test. Here you have the option of changing several different parameters to examine their effects on power.
• Power when you aren't satisfied to just reject the null. This is a document that challenges how we think about power when we are working in fields in which the mere reject of a strict null hypothesis is not sufficient.It raises some interesting questions, and does not present definitive answers. If I were one of my graduate students, I'd think about this one before the next exam on power.
• A neat, clean, and free program to calculate power for several selected tests. It offers instructors a good way to illustrate factors controlling power, and provides students the opportunity to play around with power calculations. I expect that the authors will continue to add test statistics to their list for which power can be calculated.
• A lab exercise investigating power (as well as t and correlation) is useful. It assumes SPSS, but the program could be rewritten for a different software package.
• A great example about the practical implications of power analysis can be found at Utts-Salaries.html.Utts presents an example of a situation where, because of low power, an nonsignificant difference can still be important.

Return to Dave Howell's Statistical Home Page  

University of Vermont Home Page  

Last revised: 7/11/98