A sample proportion, written as p-hat, is the fraction of a sample that has a particular characteristic. If you take many random samples from the same population, the value of p-hat will change from sample to sample. The sampling distribution of p-hat describes this variation and helps us predict how close a sample result is likely to be to the true population proportion.
This idea is central to polls, quality control, medical studies, and any situation where a percentage is estimated from data.
When the sample is random, independent, and large enough, the sampling distribution of p-hat is approximately normal. Its center is the true population proportion p, and its spread is measured by the standard error sqrt(p(1 - p)/n). Larger samples make the distribution narrower because random variation is reduced.
The normal model allows students to compute probabilities and build confidence intervals using z-scores.
Understanding Statistics: The Sampling Distribution of a Proportion
A useful way to picture this idea is to imagine one population sitting still while a researcher repeatedly draws fresh samples from it. Each sample gives one percentage. The collection of those percentages forms a new distribution.
It is not a distribution of people, products, or survey answers. It is a distribution of a statistic. Some sample results land above the population percentage and some land below it.
Random sampling makes these errors balance out over many repetitions. A single sample does not reveal where it sits in that full collection, but the sampling model describes the kind of variation that produced it.
The word random matters more than many students first expect. A larger sample reduces random noise, but it cannot repair a biased method. Suppose a school survey is sent only to students who join an online club.
Even if thousands respond, the result may not represent every student at the school. People who choose to respond can differ from people who ignore the survey. Similar problems occur when a poll misses households, uses leading wording, or surveys friends of the researcher.
The sampling distribution describes chance variation after the sampling process is fair. It does not account for mistakes built into the process.
Independence is another important detail. One person’s selection should not strongly affect another person’s chance of being selected. This can fail when data come from families, classrooms, teams, or neighbourhoods where people tend to share traits.
Sampling too large a fraction of a small population can create dependence as well. The ten percent guideline is used when sampling without replacement.
It keeps the sample small enough relative to the population that each draw has little effect on the next one. When observations are connected, the actual variation can be wider than the model predicts.
The bell-shaped approximation is not automatic. A proportion can only change in steps. With a sample of ten, it can move by one tenth at a time, so its distribution is fairly rough.
If the characteristic is rare, many samples may contain no successes at all. That creates a lopsided shape rather than a smooth bell shape.
Checking that the expected numbers of successes and failures are both at least ten helps avoid this problem. For small samples or rare events, a binomial calculation or a computer simulation gives a more reliable picture than a normal model.
In real studies, the true population proportion is usually unknown, which is the reason for collecting a sample in the first place. Researchers often use the observed proportion to estimate the amount of random error. This estimate supports confidence intervals and tests of claims.
A poll’s margin of error describes sampling variability under its assumptions, not every possible source of error. When learning this topic, keep the population proportion separate from the sample proportion. One is a fixed but usually hidden value.
The other changes whenever a new sample is drawn. That distinction prevents many common mistakes.
Key Facts
- Sample proportion: p-hat = x/n, where x is the number of successes and n is the sample size.
- Mean of the sampling distribution: mu_p-hat = p.
- Standard error of p-hat: sigma_p-hat = sqrt(p(1 - p)/n).
- Normal approximation condition: np >= 10 and n(1 - p) >= 10.
- Independence condition: observations should be independent, often checked with n <= 0.10N when sampling without replacement.
- Z-score for a sample proportion: z = (p-hat - p)/sqrt(p(1 - p)/n).
Vocabulary
- Sample proportion
- The fraction of individuals in a sample that have the characteristic being studied.
- Sampling distribution
- The probability distribution of a statistic over many random samples of the same size from the same population.
- Population proportion
- The true fraction p of the entire population that has the characteristic of interest.
- Standard error
- The standard deviation of a sampling distribution, measuring how much a statistic typically varies from sample to sample.
- Normal approximation
- The use of a bell-shaped normal model to describe a sampling distribution when sample size conditions are met.
Common Mistakes to Avoid
- Using p-hat instead of p in the standard error formula for a probability calculation. The sampling distribution under a claimed population proportion uses sigma_p-hat = sqrt(p(1 - p)/n).
- Forgetting to check np >= 10 and n(1 - p) >= 10. Without enough expected successes and failures, the sampling distribution may be skewed and the normal model may be inaccurate.
- Thinking that p-hat is always equal to p. A sample proportion is a random statistic, so it varies from sample to sample even when the population proportion stays fixed.
- Ignoring independence when sampling without replacement. If the sample is more than 10% of the population, the standard error formula may overstate or understate the true variability.
Practice Questions
- 1 A population has p = 0.40, and random samples of size n = 100 are taken. Find the mean and standard error of the sampling distribution of p-hat.
- 2 A company claims that 8% of its products are defective. For a random sample of n = 200 products, find the probability that p-hat is greater than 0.11 using the normal approximation.
- 3 A survey takes a random sample of 25 people from a population where p = 0.04. Explain whether a normal model for the sampling distribution of p-hat is appropriate, and justify your answer using the success-failure condition.