Sample Size Determination and Practical Considerations

TL;DR

Determining sample size is crucial in statistics, impacting whether you use a z-distribution (large samples, known $\sigma$) or t-distribution (small samples, unknown $\sigma$). A larger sample size generally leads to a more precise estimate of the population mean, as shown by a smaller standard error. Adequate sample size ensures reliable inferences and is often considered to be at least 30, though more may be needed for skewed data.

1. The Mental Model

Think of sample size as the amount of information you collect to understand a larger group. More information generally gives you a clearer picture, reducing uncertainty about the true population characteristics you're trying to measure.

2. The Core Material

When estimating population parameters, your sample size ($n$) significantly influences the statistical method you use and the precision of your results.

Z-Distribution vs. T-Distribution

The choice between using a z-distribution or a t-distribution for interval estimation of a population mean largely depends on two factors: the population standard deviation ($\sigma$) and your sample size ($n$).

• Z-Distribution (for population mean):

  • Used when the population standard deviation $\sigma$ is known.
  • Applied when the sample size is large (typically $n \geq 30$).
  • Its shape is symmetrical, bell-shaped, and fixed.

• T-Distribution (for population mean):

  • Used when the population standard deviation $\sigma$ is unknown.
  • Applied when the sample size is small ($n < 30$).
  • You use the sample standard deviation $s$ instead of $\sigma$.
  • Its shape is also bell-shaped and symmetric, but it's wider and has heavier tails than the z-distribution. As the sample size increases, the t-distribution approaches the z-distribution.

Relationship Between Sample Size and Sampling Distribution of $\bar{x}$

The sampling distribution of $\bar{x}$ is the probability distribution of all possible values of the sample mean $\bar{x}$.
* Effect on Standard Deviation of $\bar{x}$ ($\sigma_{\bar{x}}$):
* The standard deviation of the sample mean, also called the standard error, is calculated as $\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}$. For finite populations, it's $\sigma_{\bar{x}} = \sqrt{\frac{N-n}{N-1}} \frac{\sigma}{\sqrt{n}}$.
* A key relationship: Whenever the sample size ($n$) is increased, the standard error of the mean ($\sigma_{\bar{x}}$) is decreased. This means your sample mean ($\bar{x}$) is expected to be closer to the population mean ($\mu$).
* Effect on Expected Value of $\bar{x}$:
* The expected value of $\bar{x}$ is equal to the population mean, $E(\bar{x}) = \mu$, regardless of the sample size.

Central Limit Theorem (CLT)

The CLT is crucial when the population distribution isn't normal.
* It states that in selecting random samples of size $n$ from a population, the sampling distribution of the sample mean $\bar{x}$ can be approximated by a normal distribution as the sample size becomes large.
* This is why a sample size of $n \geq 30$ is often considered "adequate," because even if the original population isn't normal, the distribution of sample means will start looking normal.

graph TD
    A["Population (Any Distribution)"] --> B{Sample Size (n)};
    B -- "n < 30" --> C["Population Standard Deviation (σ)?"];
    C -- "σ Known" --> D["Use Z-Distribution"];
    C -- "σ Unknown" --> E["Use T-Distribution"];
    B -- "n >= 30 (Large)" --> F["Sampling Distribution of x̄ approximates Normal"];
    F --> G{"Population Standard Deviation (σ)? For Large n"};
    G -- "σ Known" --> D;
    G -- "σ Unknown (use s)" --> D;
    D -- "Interval Estimate" --> H["More Precise as n Increases"];
    E -- "Interval Estimate" --> H;

Adequate Sample Size

• In most applications, a sample size of $n \geq 30$ is considered adequate.
• However, if the population distribution is highly skewed or contains outliers, a sample size of 50 or more is recommended to ensure the sampling distribution of the mean is sufficiently normal.
• Larger samples, like the IQRA University example switching from $n=100$ to $n=900$, decrease $\sigma_{\bar{x}}$ and lead to more precise probability estimates.

3. Worked Example

Let's use the IQRA University example with the impact of changing sample size on $\sigma_{\bar{x}}$.

Suppose the population standard deviation ($\sigma$) of subject scores is 87.
* When $n = 30$, the standard error of the mean is:
$\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}} = \frac{87}{\sqrt{30}} \approx \frac{87}{5.477} \approx 15.884$
* Now, suppose IQRA University increases the sample size to $n = 900$ (as mentioned in the source for a different context). The standard error becomes:
$\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}} = \frac{87}{\sqrt{900}} = \frac{87}{30} = 2.9$

Notice how much the standard error decreased from approximately 15.884 to 2.9 when the sample size increased from 30 to 900. This smaller standard error indicates that the sample mean is much more likely to be closer to the true population mean with the larger sample.

4. Key Takeaways

• Your choice of z-distribution or t-distribution for mean interval estimation depends on whether you know the population standard deviation and your sample size.
• The t-distribution is wider and has heavier tails than the z-distribution, especially for small sample sizes.
• Increasing sample size always decreases the standard error of the mean ($\sigma_{\bar{x}}$), making your estimate more precise.
• The Central Limit Theorem ensures that the sampling distribution of the sample mean becomes approximately normal for large sample sizes, even if the population isn't normal.
• A sample size of $n \geq 30$ is generally considered adequate, but more may be needed for skewed populations or those with outliers.

Common Mistakes to Avoid:
* Using the z-distribution when the population standard deviation is unknown and the sample size is small ($n < 30$).
* Assuming the population distribution is normal just because the sample size is large (the population doesn't have to be normal, only the sampling distribution of the mean becomes normal due to CLT).
* Ignoring the recommendation for larger sample sizes (e.g., $n \geq 50$) when dealing with highly skewed data or outliers.
* Confusing the sample standard deviation ($s$) with the population standard deviation ($\sigma$).

5. Now Try It

Given a population with a standard deviation ($\sigma$) of 50, calculate the standard error of the mean ($\sigma_{\bar{x}}$) for sample sizes of $n=25$, $n=100$, and $n=400$. What do you observe about the standard error as the sample size increases?
What success looks like: You should see the standard error systematically decreasing as the sample size grows, demonstrating the core relationship.