Confidence Intervals for Small Samples
What you'll learn: Why constructing confidence intervals with fewer than 30 observations requires special care and different assumptions.
Why Small Samples Are Different
When you have a small sample (typically n < 30), the Central Limit Theorem doesn't guarantee that your sampling distribution is approximately normal. This makes confidence interval construction trickier—you can't rely on large-sample approximations.
Special Considerations
The t-Distribution is Essential
With small samples, you must use the t-distribution instead of the normal (Z) distribution. The t-distribution has heavier tails, which accounts for the extra uncertainty in your estimate. As you've learned, degrees of freedom (df = n - 1) determine the exact shape—smaller samples have heavier tails.
Population Normality Assumption
Here's the critical requirement: your underlying population should be approximately normally distributed. With large samples, the CLT compensates for non-normality. With small samples, you don't have that safety net. If your data is heavily skewed or has outliers, your confidence interval may not be trustworthy.
Wider Intervals Are Normal
Small samples naturally produce wider confidence intervals because:
- The t-critical value is larger than the Z-critical value
- Standard error is estimated with less precision
- More uncertainty = wider margin of error
Think of it like making a guess with limited information—you need to cast a wider net to be confident you've captured the truth.
Practical Checks
Before constructing a small-sample confidence interval, visually inspect your data with a histogram or Q-Q plot to verify approximate normality. Even mild departures can affect validity.
Key Takeaway: With samples smaller than 30, always use the t-distribution and verify that your data comes from an approximately normal population—without these precautions, your confidence interval may be unreliable.