Sample Size Impact

Even with a population of highly variable incomes, a sample of just 30 individuals can yield a surprisingly normal distribution of sample means. This phenomenon is a cornerstone of statistical inference, allowing us to draw reliable conclusions about large populations from smaller subsets. This lesson explores how and why increasing your sample size transforms the behavior of sample means, making them more predictable and precise.

When you take a single, small sample from a population, its mean might not accurately reflect the true population mean. Imagine surveying only five people about their income; their average might be very different from the overall average income of the city. This variability means that if you were to take many such small samples, their means would likely be spread out, making it difficult to pinpoint the true population parameter with confidence.

Looking at the distribution for $n=5$, we observe a shape that still clearly resembles the original skewed population. The peak is shifted to the left, and a long tail extends to the right. This wide spread signifies that individual sample means can vary considerably from one sample to the next. Such high variability makes it challenging to use any single sample mean as a reliable estimate for the true population mean.

With a moderate sample size of $n=30$, a significant transformation occurs. The sampling distribution of the mean now appears much more bell-shaped and symmetric, closely approximating a normal distribution. Crucially, the spread of the distribution has noticeably tightened around the population mean. This reduction in variability means that individual sample means are now much closer to the true population mean, making them more reliable estimators.

When the sample size increases further to $n=100$, the sampling distribution becomes even narrower and more concentrated around the population mean. It is now almost perfectly normal. This extreme reduction in spread means that nearly all sample means will be very close to the true population mean. Such large samples provide highly reliable and precise estimates, minimizing the impact of random sampling variability.

To formally quantify the spread of the sampling distribution of the mean, we use the standard error (SE). The standard error is essentially the standard deviation of the sample means. A smaller standard error indicates that the sample means are clustered more tightly around the population mean, implying greater precision in our estimates. It directly measures how much sample means are expected to vary from the true population mean due to random sampling.

The formula for standard error reveals a critical relationship: it is inversely proportional to the square root of the sample size, $\sqrt{n}$. This means that to halve the standard error (and thus double the precision), you need to quadruple the sample size. While increasing $n$ always reduces the standard error, the returns diminish. For example, going from $n=100$ to $n=400$ reduces the standard error by the same factor as going from $n=1$ to $n=4$.

A widely used heuristic in statistics is that a sample size of $n \ge 30$ is generally considered 'large enough' for the Central Limit Theorem to apply. This means that even if the parent population is not normally distributed, the sampling distribution of its means will be approximately normal once the sample size reaches 30 or more. This guideline is a practical rule of thumb, but it's important to remember it's a heuristic; for extremely skewed populations, a larger sample size might be needed to achieve a truly normal sampling distribution.