Parameter vs statistic: how to tell them apart

Q: Is p a parameter or a statistic?

The population proportion p is a parameter, so it describes a whole population and is usually unknown. The value you compute from a sample is the sample proportion p-hat, which is a statistic that estimates p. The hat is the signal that you are looking at a sample result.

Q: Is the sample mean a parameter or a statistic?

The sample mean x-bar is a statistic, because you compute it from the data in your sample. The matching parameter is the population mean μ, the average over the entire population, which you usually cannot measure directly. So x-bar is your best single estimate of μ.

By Jude Wallis · Published July 28, 2026

A parameter describes an entire population (like the mean mu or proportion p) and is usually unknown. A statistic is computed from a sample (like x-bar or p-hat) and estimates the matching parameter. Population pairs with parameter; sample pairs with statistic.

AP Statistics: Unit 3 (topics 3.1 Estimators). Parameters and statistics are the language of estimation in the Fall 2026 AP Statistics course. Topic 3.1 (Estimators) in Unit 3 defines a sample statistic as a point estimator of the matching population parameter and asks you to justify whether an estimator is unbiased. The same parameter idea also underlies Unit 2 topic 2.9, Parameters of Random Variables.

Parameter vs statistic: the short answer

A parameter is a number that describes an entire population, the complete group you actually care about. A statistic is a number computed from a sample, the smaller subset you were able to measure. The two travel in pairs: for almost every parameter there is a matching statistic that estimates it. The population mean $\mu$ (mu) is a parameter, and the sample mean $\bar{x}$ (x-bar) is the statistic that estimates it.

The reason the distinction matters is that you almost never get to see a parameter directly. Populations are usually too large to measure in full, so the parameter stays unknown and you use the statistic from your sample as your best estimate of it. Everything you do later in inference, from confidence intervals to hypothesis tests, is an attempt to reason from a known statistic back to an unknown parameter.

The symbol table: parameter and matching statistic

Here is the pairing for the quantities you meet most often. Read each row as "this population parameter is estimated by this sample statistic."

Quantity	Population parameter	Sample statistic
Mean	$\mu$ (mu)	$\bar{x}$ (x-bar)
Standard deviation	$\sigma$ (sigma)	$s$
Variance	$\sigma^2$	$s^2$
Proportion	$p$	$\hat{p}$ (p-hat)
Correlation	$\rho$ (rho)	$r$

Two patterns make the table easier to hold in your head. Parameters usually get Greek letters, such as $\mu$ , $\sigma$ , and $\rho$ , while their statistics get Roman letters or a hat, such as $\bar{x}$ , $s$ , and $\hat{p}$ . The proportion is the common exception, where the population value $p$ is already a Roman letter and the sample value adds a hat to become $\hat{p}$ . The same idea extends to regression, where a sample least-squares line has slope $b$ and intercept $a$ estimating an unknown population slope and intercept.

The same number can be a parameter or a statistic

Whether a number is a parameter or a statistic depends on the group it is meant to describe, not on how you calculated it. Suppose you measure the mean height of all 30 students in one classroom. If that classroom is the population you care about, the mean is a parameter, because you measured the entire group. If instead those 30 students are a sample meant to represent the whole school, the identical number is now a statistic that estimates the school-wide mean you did not measure.

So before you label a number, pin down the population first. Ask what complete group the study is really about, then check whether the number came from that whole group or only from a sample of it.

Why the parameter is usually unknown

A parameter is a fixed number. The population has some exact mean and some exact proportion right now, and those values do not change just because you did not measure them. What makes a parameter hard is access, not existence: to know $\mu$ for every adult in a country you would have to measure every adult, which is rarely possible.

A statistic is different because you compute it from data you actually collected, so its value is known to you the moment you have the sample. Take a new sample and the statistic will usually come out a little different, since it depends on which individuals happened to land in your sample. This sample-to-sample variation is the whole subject of a sampling distribution, and its spread is measured by the standard error.

How to tell them apart in one question

When a number appears in a problem, ask one question: does it describe the whole population, or only the sample that was collected? If it describes the whole group, it is a parameter. If it comes from the sample in hand, it is a statistic.

Two spelling hooks make this automatic. Population and parameter both start with p, and sample and statistic both start with s, so the words that go together share a first letter. It also helps to remember that a statistic is something you can calculate right now from your data, while a parameter is usually the unknown you are trying to estimate.

Parameters, statistics, and estimators on the AP exam

This pairing has its own place in the course. In Unit 3, topic 3.1 (Estimators), a sample statistic is called a point estimator of the matching population parameter, so $\hat{p}$ is the point estimator of $p$ and $\bar{x}$ is the point estimator of $\mu$ .

The course also asks you to judge whether an estimator is unbiased. An estimator is unbiased if, on average over all possible samples, it neither overestimates nor underestimates the parameter. The sample mean $\bar{x}$ is an unbiased estimator of $\mu$ , and the sample proportion $\hat{p}$ is an unbiased estimator of $p$ , which is one reason they are the default statistics you report.

The $n-1$ in the sample standard deviation formula is part of this same idea: dividing by $n-1$ instead of $n$ makes the sample variance $s^2$ an unbiased estimator of the population variance $\sigma^2$ . You can practice the mechanics on the standard deviation calculator or work through one by hand.

Common mistakes to avoid

A handful of mix-ups cause most of the lost points.

Do not call a sample result a parameter. A poll's 61% support is a statistic ( $\hat{p}$ ) because it came from the people polled, not from every voter.
Do not assume a parameter is known. In real problems $\mu$ , $\sigma$ , and $p$ are the unknowns you estimate, which is exactly why you took a sample.
Do not mix the symbols. Writing $\bar{x}$ when you mean $\mu$ , or $p$ when you mean $\hat{p}$ , changes the meaning of a whole calculation.
Do not confuse the sample size $n$ with a statistic. It is a known feature of your sample, not an estimate of any population quantity.
Do not treat the sample standard deviation $s$ and the population standard deviation $\sigma$ as interchangeable, since one is computed with $n-1$ and estimates the other.

Classify each quantity in a voter poll

A political scientist knows from a national registry that the true proportion of all registered voters who support a measure is 0.64. A newspaper then polls 1,200 randomly chosen registered voters, finds that 61% of them support the measure, and reports that the mean age of the 1,200 people polled is 47 years. Label each quantity as a parameter or a statistic and give its symbol.

Find the group each number describes. The value 0.64 covers all registered voters, which is the population, so it is a parameter. Its symbol is the population proportion $p$ , giving $p = 0.64$ .
The 61% support comes only from the 1,200 people polled, which is the sample, so it is a statistic. Its symbol is the sample proportion $\hat{p}$ , giving $\hat{p} = 0.61$ .
The mean age of 47 years is also computed from the 1,200 polled, so it is a statistic too. Its symbol is the sample mean $\bar{x}$ , giving $\bar{x} = 47$ years.
The 1,200 is the number of people in the sample, so it is the sample size $n = 1200$ . It is a known feature of the sample, not an estimate of any population value, so it is neither a parameter nor a statistic in this sense.

Parameter: $p = 0.64$ (all registered voters). Statistics: $\hat{p} = 0.61$ and $\bar{x} = 47$ years (both from the 1,200 polled). The 1,200 is the sample size $n$ , a feature of the sample rather than an estimate. The statistic $\hat{p} = 0.61$ is the poll's estimate of the parameter $p = 0.64$ .

Match a sample proportion to its population proportion

A school has 1,800 students, and school records show that 1,080 of them plan to attend a four-year college. A counselor takes a random sample of 50 students and finds that 32 of them plan to attend a four-year college. Identify the parameter and the statistic, compute each, and compare.

The 1,800 students are the whole population, and 1,080 of them plan to attend, so the population proportion is a parameter: $p = \frac{1080}{1800}$ .
Compute it: $\frac{1080}{1800} = 0.60$ . So $p = 0.60$ , the true proportion for the entire school.
The 50 sampled students give a sample proportion, which is a statistic: $\hat{p} = \frac{32}{50}$ .
Compute it: $\frac{32}{50} = 0.64$ . So $\hat{p} = 0.64$ , the estimate from this one sample.
Compare the two: $\hat{p} - p = 0.64 - 0.60 = 0.04$ . The statistic missed the parameter by 0.04, an ordinary amount of sampling variability for a sample this size.

Parameter: $p = 0.60$ (all 1,800 students). Statistic: $\hat{p} = 0.64$ (the 50 sampled students). The statistic estimates the parameter and here overshoots it by 0.04, which is the kind of sample-to-sample difference you expect.

Sort mu and sigma from x-bar and s

A bottling machine is set so that, over a long production run, the mean fill is $\mu = 500$ mL with standard deviation $\sigma = 4$ mL. An inspector pulls a random sample of 5 bottles and measures 498, 502, 501, 499, and 505 mL. Identify which numbers are parameters and which are statistics, then compute the sample mean and sample standard deviation.

The machine's long-run values describe the whole population of bottles, so they are parameters: the population mean $\mu = 500$ mL and the population standard deviation $\sigma = 4$ mL.
The five measured bottles are a sample, so the numbers you compute from them will be statistics: the sample mean $\bar{x}$ and the sample standard deviation $s$ .
Sample mean: $\bar{x} = \frac{498 + 502 + 501 + 499 + 505}{5} = \frac{2505}{5} = 501$ mL.
Find each deviation from 501: $-3, +1, 0, -2, +4$ . Square them: $9, 1, 0, 4, 16$ , which sum to 30.
Sample standard deviation uses $n-1 = 4$ : $s = \sqrt{\frac{30}{4}} = \sqrt{7.5} \approx 2.74$ mL.

Parameters: $\mu = 500$ mL and $\sigma = 4$ mL (the whole production run). Statistics: $\bar{x} = 501$ mL and $s \approx 2.74$ mL (the sample of 5). Each statistic estimates the parameter directly above it, and $s$ is computed with $n-1$ , which marks it as a sample standard deviation.

Frequently asked questions

Is p a parameter or a statistic?

The population proportion $p$ is a parameter, so it describes a whole population and is usually unknown. The value you compute from a sample is the sample proportion $\hat{p}$ , which is a statistic that estimates $p$ . The hat is the signal that you are looking at a sample result.

Is the sample mean a parameter or a statistic?

The sample mean $\bar{x}$ is a statistic, because you compute it from the data in your sample. The matching parameter is the population mean $\mu$ , the average over the entire population, which you usually cannot measure directly. So $\bar{x}$ is your best single estimate of $\mu$ .

Do a statistic and its parameter have to be equal?

No. A statistic almost always differs from the parameter it estimates, because a sample is only part of the population and shifts from sample to sample. That gap is the reason you attach a margin of error or standard error when you report a statistic.