Sampling Distributions for
Differences in Sample Proportions
Learn how the distribution of \(\hat{p}_1 - \hat{p}_2\) behaves, when it's approximately normal, and how to calculate its mean and standard deviation.
What Is a Sampling Distribution?
When we take a random sample of size \(n\) from a population and compute a statistic (like a sample proportion \(\hat{p}\)), that statistic varies from sample to sample. The sampling distribution describes all possible values the statistic can take and how likely each value is.
From One Proportion to Two
You already know that \(\hat{p}\) for a single sample has a sampling distribution centered at \(p\). Now imagine two independent populations with true proportions \(p_1\) and \(p_2\). We draw a sample of size \(n_1\) from the first and \(n_2\) from the second, computing \(\hat{p}_1\) and \(\hat{p}_2\).
Why Does This Matter?
Comparing two proportions is one of the most common tasks in statistics: Does a new drug work better than a placebo? Do men and women differ in voting preference? Is one manufacturing process more reliable than another?
The Core Result: Mean and Standard Deviation
If samples are drawn independently from two populations, the sampling distribution of \(\hat{p}_1 - \hat{p}_2\) has:
When Is It Approximately Normal?
The sampling distribution of \(\hat{p}_1 - \hat{p}_2\) is approximately normal when all four of the following conditions are met:
Visualizing \(\hat{p}_1 - \hat{p}_2\)
Independence vs. Equal Proportions
Two conditions are easy to confuse but serve very different purposes:
| Independent Samples | Equal Proportions (\(p_1 = p_2\)) | |
|---|---|---|
| What it means | The two samples don't influence each other | The null hypothesis assumes no difference |
| When needed | Always: for the formula to be valid | Only in a hypothesis test under \(H_0\) |
| Effect on formula | Allows adding variances | Lets us use a pooled \(\hat{p}\) |
The 10% Condition
For the standard deviation formula to be accurate, each sample must be drawn without replacement from a population that is at least 10 times the sample size:
This ensures that sampling without replacement behaves approximately like sampling with replacement, so the observations within each sample remain nearly independent.
How to Apply the Formula
Example 1: Drug Trial
A clinical trial tests a new drug. In the treatment group (\(n_1 = 200\)), the true success rate is \(p_1 = 0.60\). In the placebo group (\(n_2 = 150\)), \(p_2 = 0.45\). Describe the sampling distribution of \(\hat{p}_1 - \hat{p}_2\).
Example 1: Conclusion
Example 2: Voter Preference
A pollster surveys \(n_1 = 500\) voters in City A and \(n_2 = 400\) in City B. Suppose \(p_1 = 0.52\) and \(p_2 = 0.48\). Find \(\mu\) and \(\sigma\) for \(\hat{p}_1 - \hat{p}_2\).
Example 3: Finding a Probability
Using Example 2, what is the probability that \(\hat{p}_1 - \hat{p}_2 < 0\) (the poll shows City B ahead)?
Common Mistakes
Summary: One Proportion vs. Difference
| Single \(\hat{p}\) | \(\hat{p}_1 - \hat{p}_2\) | |
|---|---|---|
| Mean | \(p\) | \(p_1 - p_2\) |
| Std Dev | \(\sqrt{\dfrac{p(1-p)}{n}}\) | \(\sqrt{\dfrac{p_1(1-p_1)}{n_1} + \dfrac{p_2(1-p_2)}{n_2}}\) |
| Shape | Approx Normal if \(np \geq 10\), \(n(1-p) \geq 10\) | Approx Normal if all four Large Counts \(\geq 10\) |
| Key Condition | \(n \leq 0.10 N\) | Both \(n_1 \leq 0.10 N_1\) and \(n_2 \leq 0.10 N_2\) |
Question 1 of 5
What is the mean of the sampling distribution of \(\hat{p}_1 - \hat{p}_2\)?
Question 2 of 5
Why do we add the variance terms \(\dfrac{p_1(1-p_1)}{n_1}\) and \(\dfrac{p_2(1-p_2)}{n_2}\) rather than subtract them?
Question 3 of 5
Samples of \(n_1 = 80\) and \(n_2 = 60\) are drawn. Which set of conditions must be met for the sampling distribution of \(\hat{p}_1 - \hat{p}_2\) to be approximately normal?
Question 4 of 5
If \(p_1 = 0.70\), \(p_2 = 0.50\), \(n_1 = 100\), and \(n_2 = 100\), what is the standard deviation of \(\hat{p}_1 - \hat{p}_2\)?
Question 5 of 5
A researcher wants to compare proportions from two independent populations. The 10% condition requires: