Guide to Statistics: "Supporting Statistics in Medicine;"

Print this page

10 Sample size

When planning a study it is important to have a good idea of how many individuals are needed to take part. The justification for a chosen sample size can range from a formal calculation, to what is feasible. A sample size calculation is also called a power calculation as it involves answering the question "How many patients are needed in order to detect a particular size of difference for a pre-determined power (usually at least 80%) at a particular level of statistical significance (usually 5%)?". In order to calculate the size of the sample required, several quantities need to be known or estimated a priori as listed below:

• The number of groups to be compared
• The outcome variable (and its distribution)
• Standard deviation of the outcome variable
• Minimum Clinically Important Difference and the size of difference likely to be detected. The latter should be at least as great as the former. If the size of difference that is likely to be detected is smaller than the minimally important difference it is arguable whether it is worth doing the study at all.
• The acceptable level for the type 1 error rate $$\alpha$$. Usually set at 5%
• The power $$1-\beta$$. Usually set at either 80% or 90%.

In addition, it is worth considering drop-outs when calculating sample size as the number calculated is the number necessary for analysis, the analysable sample. To account for potential dropouts this should be scaled up in order to give the sample size to be recruited.

Contents