Power for a One-sample T-test or Paired T-test

• 1). Choose a minimum difference you want to be able to find. For example, if you are testing if the weight of your sample of people is different from 160 pounds, you might choose a difference of 10 pounds as being important.
  
  
• 2). Go to http://www.stat.uiowa.edu/~rlenth/Power/index.html and choose "one sample t-test or paired t" from the menu on the left.
  
  
• 3). Decide which quantities to enter. You must enter four of the five quantities, and solve for the fifth. For justifying sample size, you will probably want to enter power, alpha, sample size and sigma (standard deviation).
  
  
• 4). Enter values. You can enter specific values by clicking on the small squares to the right of the scales. n is the sample size you wish to justify. Power and alpha are usually set by convention in a particular field (either 0.8 or 0.9 for power and either 0.05 or 0.01 for alpha). The standard deviation can be calculated from your sample or estimated from the literature. You will almost always want two-tailed tests. You can then solve for "true mu - mu0" (the minimum difference).
  
  
• 5). Choose "solve for effect size" from that pull-down menu. If you have sigma = 10, n = 100, power = 0.8, alpha = 0.05, two-tailed, then the true mu - mu0 is 4.80, meaning you will have power of 0.8 to detect a difference of 4.8 pounds.
  
  

Justifying Sample Size for an Independent t-test

• 1). Choose a minimum difference you want to be able to find. For example, if you are testing if the weight of your sample of people is different from 160 pounds, a difference of 10 pounds might be selected as being important.
  
  
• 2). Go to http://www.stat.uiowa.edu/~rlenth/Power/index.html and choose "two sample t-test" from the menu on the left.
  
  
• 3). Decide which quantities to enter. You must enter six of the seven quantities, and solve for the quantities. For justifying sample size, you will probably want to enter power, alpha, sample size for each sample and sigma (standard deviation) for each sample.
  
  
• 4). Enter values. You can enter specific values by clicking on the small squares to the right of the scales. Power and alpha are usually set by convention in a particular field (either 0.8 or 0.9 for power and either 0.05 or 0.01 for alpha). n1 and n2 will be your sample sizes. The standard deviation for each sample can be calculated from your samples or estimated from the literature. You will almost always want two-tailed tests. You can then solve for "true mu - mu0" (the minimum difference).
  
  
• 5). Choose "solve for effect size" from that pull-down menu.