Sample Size and Normal Approximation
What you'll learn: You'll discover how the sample size affects the speed at which sample means converge to a normal distribution under the Central Limit Theorem.
The Core Idea
The Central Limit Theorem tells us that sample means become approximately normally distributed as sample size grows—but how large does n need to be for this approximation to work well? The answer depends on the shape of the original population distribution.
When you're sampling from a population that's already close to normal, even small samples (n = 5 or 10) may produce sample means that are approximately normal. However, if the underlying distribution is heavily skewed or has unusual features (like extreme outliers), you might need n = 30, 50, or even larger before the normal approximation becomes reliable.
Why Sample Size Matters
Think of the CLT as a smoothing process: each additional observation you add to your sample helps "average out" the quirks of the original distribution. The more extreme or irregular your starting distribution, the more observations you need to smooth it into a bell curve.
Rule of thumb: For moderately skewed distributions, n ≥ 30 is often sufficient. For symmetric distributions, n ≥ 15–20 may work well. For highly skewed or heavy-tailed distributions, you may need n ≥ 50 or more.
Practical Implications
When applying the normal approximation (for confidence intervals, hypothesis tests, or probability calculations), always consider:
- The shape of your population distribution
- Your sample size relative to that shape
- Whether the approximation error is acceptable for your application
Key Takeaway: The rate at which sample means approach normality depends critically on both the sample size n and the shape of the underlying distribution—symmetric distributions converge faster than skewed ones, making sample size selection a context-dependent decision.