Effect Size in Sample Size Planning
You'll learn how to use anticipated effect sizes to calculate the sample size you need to detect meaningful differences with adequate statistical power.
Why Effect Size Matters for Planning
When designing a study, you can't just guess how many participants you need. Your sample size should be large enough to detect the effect you care about—if it exists—with high probability. This is where effect size becomes your planning tool.
Remember, statistical power is the probability of correctly rejecting a false null hypothesis. But power depends on how big the true effect is. A tiny effect requires a much larger sample to detect than a large effect.
Using Effect Size to Determine Sample Size
Before collecting data, ask yourself: "What's the smallest effect that would be practically meaningful?" This becomes your anticipated effect size.
For example, if you're testing whether a new teaching method improves test scores, you might decide that only improvements of Cohen's d ≥ 0.5 (a medium effect) would justify the cost of implementation. You'd then calculate: "How many students do I need to have 80% power to detect d = 0.5 at α = 0.05?"
The formula connects four quantities:
- Desired power (typically 0.80)
- Significance level (α, typically 0.05)
- Anticipated effect size (from theory, prior studies, or practical considerations)
- Required sample size (what you're solving for)
The Planning Process
- Specify your anticipated effect size based on domain knowledge or pilot data
- Choose your desired power level (usually 80% or 90%)
- Set your significance level
- Use power analysis formulas or software to calculate required n
Key Takeaway: Effect size translates "How big a difference matters?" into "How many observations do I need?"—making it the bridge between scientific importance and statistical feasibility in study design.