Evidence guide
Understanding Sample Size
Sample size affects precision and the chance of detecting effects, but a large dataset is not automatically unbiased.
Small studies
They can be appropriate for rare conditions or mechanistic work but tend to produce wide uncertainty and unstable effect estimates.
Large studies
They can detect tiny differences that are statistically clear but practically trivial. Systematic bias does not disappear with more participants.
The unit matters
Ten thousand images from twenty patients may still have only twenty independent human units. Clustered observations must be analyzed accordingly.
A practical reading checklist
- Was a power calculation reported?
- What is the independent unit?
- How many events occurred?
- Is the effect clinically or scientifically meaningful?