Statistics – Statistical Power

Statistical Power

Definition

\(\text{Power} = 1 - \beta\)

= probability of rejecting \(H_0\) when it is false

Interpretation

  • High power = test is good at detecting a true difference
  • Low power = high chance of missing a true difference

What affects power?

  • Sample size (n↑ → power↑)
  • Significance level (\(\alpha\)↑ → power↑)
  • True effect size (larger difference → power↑)
  • Population variance (lower variance → power↑)

Convention

Desired power: at least 80% (0.8)