Under a normal approximation with fixed variance, alpha and power, what happens to required n if the detectable effect…
StatisticsLevel: IntermediateMultiplicity and experimental power
Under a normal approximation with fixed variance, alpha and power, what happens to required n if the detectable effect halves?
Hints
- Keep effect times square root of n constant.
- Square the inverse effect ratio.
Answer
Required sample size grows approximately fourfold because the signal-to-noise ratio scales as effect×√n. Integer and finite-sample corrections can alter the exact result.
Source and licence: statsmodels — statistical power · BSD-3-Clause · answer: Quantoros editorial