In Reply: Dr Janosky correctly points out that
statistical power is a continuous quantity. Accepting the value of .80 as
adequate power is as arbitrary as accepting .05 to denote statistical significance.
Greater power and stricter standards for significance will always be preferable,
but one is always obtained at the cost of the other. Our experience has been
that many peer reviewers still adhere to a dichotomy based on 80% or 90% power.
However, no numerical value for power will be adequate if it is generated
without carefully considering the size of the clinically important effect.
The 3-tiered model we propose may help investigators determine such effects
and calculate their power accordingly.
Halpern SD, Karlawish JHT, Berlin JA. The Ethics of Underpowered Clinical Trials—Reply. JAMA. 2002;288(17):2118-2119. doi:10.1001/jama.288.17.2118