Power and Effect Size, Decided Before the Data
Power depends on the effect size worth detecting, the variability and n, and it has to be settled beforehand rather than computed afterwards from what happened. This stop runs the calculation and treats the smallest useful effect as a scientific decision.
What a learner can do afterwards
- Compute the sample size needed for a stated effect size, variability and power
- Explain why power computed after the fact from the observed effect adds nothing
- Argue why a significant result from an underpowered study tends to overstate the effect
1 · Read
You fix power before data collection, not after. Power is your chance of catching a real effect, set by the effect size, the variability, the sample size, and the significance threshold. Small noisy studies start weak and stay weak no matter how results look afterward.
You trade the levers deliberately. Bigger effects, quieter measurement, larger samples, or looser thresholds each raise power, and each has a price. You state the smallest effect worth finding and convert it into the sample you must enroll.
You finish a study and compute power from your observed effect. Your mentor stops you: the observed effect already decided the p value, so the calculation launders the same number into new clothes. Honest power always uses an effect chosen before seeing data.
You distrust significant hits from underpowered work. Only inflated estimates clear the bar when power is low, so published winners overstate the truth on average. You run sensitivity checks across plausible effects, and you replicate with proper power.
You choose the effect first, you size the study for it, and you never compute power from what happened.
2 · Watch
Take it off screen
Where it sits
Where this leads
8 questions wait behind this lesson, each with its answer explained. Every answer feeds the sky: stars light as they are learned, and dim when it is time to come back.