Power Analysis

The planning step that asks “if a real effect of this size exists, what is the chance my experiment will actually detect it?” That chance is the statistical power (統計檢定力), and teams usually aim for 80% or more.

Power is the flip side of a Type II error (型二錯誤) — failing to find an effect that is genuinely there. Power analysis ties together sample size, effect size, variance, and significance level so you can see the trade-offs and avoid the most common experiment failure: an underpowered test that quietly returns “nothing here” simply because it was too small to see anything.