You can use Minitab to determine the sample size you need for many basic hypothesis tests, including tests that assess population means, proportions, rates, and other parameters. Choose and select the analysis you want to perform.
To calculate the amount of data you should collect for a test, or the power function of a test, you need to know:
- Standard deviation
- The power function of a test for a population mean of difference between two population means assumes that you know the population standard deviation. Because you usually don't know the value of the population standard deviation, use a historical estimate or the standard deviation of a sample. For example, you want to know whether the mean fill weight of cereal boxes is within 0.5 oz of the target (20 oz). Historically, the standard deviation of fill weights for this machine is 0.9 oz, so you use this value as the population standard deviation.
- Size of a relevant difference
- This value is the smallest difference between the true population parameter and the hypothesized value that has practical consequences for your situation. It is usually the difference that you want the hypothesis test to detect. For example, a quality expert may decide that a relevant difference between the mean width of dowels manufactured on a machine and the target width is 0.05 cm. Any difference less than 0.05 cm will not have a meaningful effect on the usage of the dowels. This difference is also known as the population effect, or just, the effect.
To determine the required sample size, you need to know the standard deviation, the size of the difference, and the target power for the test. If you have limited resources and only have a certain number of units available for your analysis, you can calculate the size of the difference the test can detect or you can determine the power associated with some practical difference you want the test to detect.
Usually, you weigh power and sample size requirements in relation to your available time and resources. The optimal power or sample size is determined by whether the incremental value of improved power is offset by the costs of obtaining additional sample units.
