Glossary
Statistical power.
Direct answer
Statistical power is the probability that a study will detect a difference that genuinely exists. It rises with the number of subjects examined, with the size of the effect expected and with the precision of the measurement. A study with low power often finds nothing although there was something to find.
Hence the most useful reading rule: finding no difference is not the same as there being no difference. Power must be planned before the start, because calculated afterwards it merely restates the result. Underpowered studies also produce exaggerated effect sizes on the occasions when they do find something by chance.
In equine research low power is the norm, because animal numbers, costs and ethical requirements limit group sizes. The remedies are more precise measurement, crossover designs comparing each animal with itself, and joint analysis across several clinics. Human medicine solves the problem far more often through large study networks.
Confidence intervalRandomised controlled trialPreregistration
The evidence letter
Once a month: new syntheses, checked studies, and what has changed in the state of the evidence.