Statistic vs Parameter: Quick Answer
A statistic is a number calculated from a sample — a subset of a population. A parameter is a number that describes an entire population. Because measuring an entire population is usually impossible, researchers use statistics to estimate unknown parameters. Statistics use Roman (Latin) letters such as x̄, s, and p̂. Parameters use Greek letters such as μ, σ, and p. The sample mean x̄ is a statistic; the population mean μ is its corresponding parameter.
Sample Statistic
- Calculated from a sample
- Value is known (you measured it)
- Changes from sample to sample
- Roman (Latin) letters
- Used to estimate the parameter
- Example: x̄ = 72.4 (sample mean)
Population Parameter
- Describes the entire population
- Usually unknown (rarely measured)
- Fixed — it does not change
- Greek letters
- The target of inference
- Example: μ = ? (population mean)
What Is a Statistic?
The word "statistic" comes from the German Statistik, which Karl Pearson and later Ronald Fisher formalized in the early twentieth century as any function of observed sample data. In modern usage, the American Statistical Association defines a statistic as any quantity derived from a sample that can be used for inference about a population.
Three properties distinguish statistics from parameters. First, a statistic is computed — you calculate it from data you actually collected. Second, it is variable — draw a different sample and you get a slightly different value of the same statistic. Third, it is observable — you can see it, compute it, and report it precisely.
Common Sample Statistics
| Statistic | Symbol | Formula | What It Measures |
|---|---|---|---|
| Sample Mean | x̄ (x-bar) | x̄ = Σxᵢ / n | Average value in the sample |
| Sample Standard Deviation | s | s = √[Σ(xᵢ − x̄)² / (n−1)] | Spread around the sample mean |
| Sample Variance | s² | s² = Σ(xᵢ − x̄)² / (n−1) | Average squared deviation |
| Sample Proportion | p̂ (p-hat) | p̂ = x / n | Fraction with the attribute of interest |
| Sample Size | n | — | Number of observations in the sample |
| Sample Correlation | r | Pearson r formula | Linear relationship between two variables |
| Sample Median | x̃ (x-tilde) | Middle ordered value | Center of the sample, resistant to outliers |
The sample mean formula uses n in the denominator. The sample variance formula uses n − 1 (called Bessel's correction) to produce an unbiased estimate of the population variance — dividing by n would systematically underestimate σ². This distinction matters when you move from describing your data to making inferences about the population. You can explore the mathematical basis in the variance guide on Statistics Fundamentals.
What Is a Parameter?
The word "parameter" derives from the Greek para (beside) + metron (measure) — literally, a "side measure" or boundary condition. In statistics, it entered common use through Jerzy Neyman's work on estimation in the 1930s. Today, every major statistics textbook from OpenStax to DeGroot & Schervish defines a parameter as a characteristic of a population, as opposed to a statistic, which characterizes a sample.
Parameters are fixed but unknown. The average height of every adult in the United States right now is a single, definite number. It does not fluctuate — it is whatever it is. The problem is that measuring every adult American is not feasible. That fixed, inaccessible value is the parameter. What the survey team measures from their 1,200-person sample is the statistic.
Common Population Parameters
| Parameter | Symbol | Formula | What It Describes |
|---|---|---|---|
| Population Mean | μ (mu) | μ = Σxᵢ / N | Average value across the entire population |
| Population Standard Deviation | σ (sigma) | σ = √[Σ(xᵢ − μ)² / N] | Spread around the population mean |
| Population Variance | σ² (sigma squared) | σ² = Σ(xᵢ − μ)² / N | Average squared deviation in the population |
| Population Proportion | p | p = X / N | True fraction with the attribute of interest |
| Population Size | N | — | Total number of individuals in the population |
| Population Correlation | ρ (rho) | Pearson ρ formula | True linear relationship in the population |
| Population Median | M or η (eta) | True middle value | Center of the population distribution |
Notice that the population variance formula divides by N (the full population size), not N − 1. Bessel's correction only applies when estimating σ² from a sample. When you have every single observation in the population, there is nothing to correct for — σ² is the exact variance, not an estimate.
Statistic vs Parameter: The Key Differences
The table below covers every meaningful dimension on which statistics and parameters differ. Memorizing these distinctions is the first step toward understanding hypothesis testing, confidence intervals, and the entire field of statistical inference.
| Dimension | Statistic | Parameter |
|---|---|---|
| Calculated from | A sample (subset of the population) | An entire population |
| Notation (mean) | x̄ (Roman) | μ (Greek) |
| Notation (SD) | s (Roman) | σ (Greek) |
| Notation (proportion) | p̂ — p-hat (Roman) | p (Greek context) |
| Notation (size) | n (lowercase) | N (uppercase) |
| Known or unknown? | Known — you computed it | Usually unknown — must be estimated |
| Fixed or variable? | Varies across different samples | Fixed (one true value exists) |
| Purpose | To estimate the parameter | The true population characteristic |
| Denominator (variance) | n − 1 (Bessel's correction) | N (no correction needed) |
| When is it used? | When a census is not feasible | Census data or full-population records |
| Examples | x̄ = 68.2 in. (survey of 500 adults) | μ = 69.0 in. (all U.S. adults, CDC data) |
Greek vs Roman Symbols: The Complete Reference
The most reliable shortcut for distinguishing statistics from parameters is the letter convention: statistics use Roman (Latin) letters; parameters use Greek letters. This convention originated with Karl Pearson, who in the late 1800s began using Greek letters for population quantities to distinguish them from the sample quantities Galton and others had been measuring.
Greek = Population Parameter. Roman = Sample Statistic. Think: "Go to the Population with Greek letters."
The population proportion is written p (a Roman letter), not a Greek letter. This is the one common exception to the Greek = parameter rule. The sample proportion is written p̂ (p-hat). The hat (^) consistently signals "this is a sample estimate," regardless of which letter is used.
Population vs Sample
Understanding the distinction between statistic and parameter starts with understanding population vs sample. Every study involves a population — the complete group of interest — and most studies work with a sample taken from it.
| Aspect | Population | Sample |
|---|---|---|
| Definition | Every individual of interest in the study | A selected subset of the population |
| Numerical summary | Parameter (μ, σ, p) | Statistic (x̄, s, p̂) |
| Typically known? | No — too large to measure | Yes — you collected the data |
| Data collection method | Census (rare and expensive) | Random sampling (common) |
| Example (students) | All 3.2 million AP Statistics students in the U.S. | 500 randomly chosen AP Statistics students |
| Size notation | N (capital) | n (lowercase) |
When a census is feasible — like the U.S. decennial census that attempts to count every resident — the resulting numbers are parameters, not statistics. The U.S. Census Bureau reports that this is why census figures are treated differently from survey estimates in federal statistics. For most research, though, a well-designed random sample and the statistics derived from it are what drive conclusions about the population. The sampling distributions page on Statistics Fundamentals explains how those statistics behave across repeated samples.
Worked Examples: Identifying Statistics and Parameters
The six examples below span education, healthcare, polling, manufacturing, finance, and sports. For each one, identify the population, the sample, the parameter, and the statistic — then read the explanation.
Example 1 — Education: Exam Scores
A school district has 8,400 students. A researcher randomly selects 200 students and finds their mean math exam score is 78.5 out of 100. The district's goal is to know the true average score for all students.
Population: All 8,400 students in the district (N = 8,400)
Sample: The 200 randomly selected students (n = 200)
Statistic: x̄ = 78.5 — the sample mean exam score. This is known because it was calculated from the 200 observed scores.
Parameter: μ = ? — the population mean exam score for all 8,400 students. This is unknown. The statistic x̄ = 78.5 is used to estimate it.
✅ x̄ = 78.5 is the statistic (from the sample). μ is the parameter (for the population). The researcher uses x̄ to estimate μ via statistical inference, for example constructing a confidence interval for the mean.
Example 2 — Healthcare: Blood Pressure Clinical Trial
A pharmaceutical company tests a new antihypertensive drug on 150 patients. The sample shows a mean systolic blood pressure reduction of 12.4 mmHg. The company wants to know the true average reduction the drug produces in all hypertensive patients.
Population: All hypertensive patients who could receive this drug (potentially millions)
Sample: The 150 trial participants (n = 150)
Statistic: x̄ = 12.4 mmHg reduction — calculated from the 150 trial observations. Also relevant: s (sample SD), which captures variability in individual patient responses.
Parameter: μ = the true mean blood pressure reduction in all patients. Unknown. The FDA requires statistical evidence — a hypothesis test — that μ > 0 before approving the drug.
✅ Statistic: x̄ = 12.4 mmHg (from 150-patient sample). Parameter: μ = true average reduction in the full population of patients. Clinical trials are the canonical example of using statistics to estimate — and test claims about — population parameters. See the hypothesis testing in clinical trials guide for a full walkthrough.
Example 3 — Election Polling
An election poll surveys 1,200 registered voters. 624 say they intend to vote for Candidate A. The pollster reports a support rate and wants to know the true proportion among all registered voters.
Population: All registered voters in the electorate (may be tens of millions)
Sample: The 1,200 surveyed voters (n = 1,200)
Statistic: p̂ = 624 / 1,200 = 0.52 or 52%. This sample proportion is the known statistic.
Parameter: p = the true proportion of all registered voters who support Candidate A. Unknown until election day — which is why polls report a margin of error alongside p̂. The margin of error reflects how far p̂ could plausibly be from p.
✅ Statistic: p̂ = 0.52 (from the 1,200-person poll). Parameter: p = true population proportion. This example illustrates why a poll with p̂ = 52% can still lose — p may be 49%. The margin of error quantifies the uncertainty between the statistic and the parameter.
Example 4 — Manufacturing Quality Control
A factory produces 50,000 bolts per day. A quality engineer randomly samples 75 bolts and measures their diameters. The mean diameter is x̄ = 12.03 mm with s = 0.08 mm. The specification requires μ = 12.00 ± 0.15 mm.
Population: All 50,000 bolts produced that day (N = 50,000)
Sample: The 75 measured bolts (n = 75)
Statistics: x̄ = 12.03 mm (sample mean diameter), s = 0.08 mm (sample standard deviation). Both are computed from the 75 measurements.
Parameters: μ = true mean diameter for all 50,000 bolts (specification: 12.00 mm), σ = true standard deviation of the batch. Both are unknown. The engineer uses a one-sample t-test to determine whether μ = 12.00 mm is supported by the sample.
✅ Two statistics here: x̄ = 12.03 mm and s = 0.08 mm. Two parameters: μ (true mean) and σ (true SD). Manufacturing quality control is a real-world application where the one-sample t-test uses sample statistics to test claims about population parameters.
Example 5 — Finance: Investment Returns
An analyst studies 36 months of returns for a mutual fund and calculates a mean monthly return of x̄ = 0.85% with a sample standard deviation of s = 2.1%. She wants to know if the fund's true average monthly return is positive.
Population: All monthly returns the fund will ever produce (a conceptual, infinite population of returns)
Sample: The 36 observed monthly returns (n = 36)
Statistics: x̄ = 0.85% and s = 2.1% — both calculated from the 36 months of data.
Parameter: μ = the true expected monthly return of the fund. Unknown. The analyst tests H₀: μ = 0 vs H₁: μ > 0 using a one-sample t-test, with df = 35.
✅ Statistic: x̄ = 0.85%, s = 2.1% (from 36 months). Parameter: μ = true average monthly return. Finance uses this framework constantly — past returns are the sample; the fund's true expected return is the parameter investors care about.
Example 6 — Sports: Free Throw Percentage
An NBA player attempted 120 free throws in the first half of a season and made 89 of them. Is the player's true free throw success rate (his career parameter) above 70%?
Population: All free throw attempts this player will ever take in his career (or the current season — defining the population precisely matters)
Sample: The 120 free throws attempted so far (n = 120)
Statistic: p̂ = 89 / 120 = 0.742 (74.2%) — the sample proportion made from the 120 attempts observed.
Parameter: p = the player's true free throw percentage over the full population of career attempts. Unknown — the season isn't over. A proportion z-test checks whether p > 0.70.
✅ Statistic: p̂ = 0.742 (from 120 observed shots). Parameter: p = true career free throw proportion. Sports analytics routinely use small samples (half a season) to estimate career-level parameters, which is why early-season statistics can be misleading predictors of long-term performance.
How Statistics Estimate Parameters: Statistical Inference
The entire field of inferential statistics rests on this fundamental relationship: a sample statistic is an estimate of an unknown population parameter. The process of drawing conclusions about parameters from statistics is called statistical inference.
There are two main inferential approaches. Estimation produces a point estimate (the statistic itself, e.g., x̄ = 14.3) or an interval estimate — a confidence interval that captures the parameter with a stated probability. Hypothesis testing uses the statistic to evaluate a specific claim about the parameter, such as H₀: μ = 12. Both are covered in depth across the core guides on Statistics Fundamentals.
μ (unknown) estimated by x̄ (known)
σ (unknown) estimated by s (known)
p (unknown) estimated by p̂ (known)
The quality of the estimate depends on three things: sample size, sampling method, and the variability in the population. A larger n reduces the standard error (SE = σ / √n), making the statistic a more precise estimate of the parameter. Random sampling ensures the statistic is unbiased — on average equal to the parameter.
The standard deviation (s or σ) describes variability in individual observations. The standard error (SE = s / √n) describes variability in the statistic — how much x̄ would vary across repeated samples. As n increases, SE decreases, so the statistic becomes a more precise estimate of the parameter. The Central Limit Theorem guarantees that x̄ is approximately normal regardless of the population shape, provided n is large enough.
Point Estimate vs Interval Estimate
| Estimation Type | What It Reports | Example |
|---|---|---|
| Point Estimate | A single statistic as the best guess for the parameter | μ ≈ x̄ = 78.5 (sample mean as estimate of population mean) |
| Interval Estimate (Confidence Interval) | A range of plausible parameter values with a stated confidence level | 95% CI for μ: (75.2, 81.8) — we're 95% confident μ lies here |
| Hypothesis Test | A decision about whether the parameter equals a specific claimed value | Test H₀: μ = 75 vs H₁: μ ≠ 75. Reject or fail to reject based on x̄ |
Parameter vs Statistic vs Estimator: The Full Picture
A complete understanding of statistical inference requires one more distinction. An estimator is the rule or formula used to compute a statistic from data. The statistic itself — the specific number that results from applying the estimator to a particular sample — is called the estimate.
| Concept | Definition | Example |
|---|---|---|
| Parameter | Fixed, unknown population characteristic | μ = true average IQ of all adults |
| Estimator | The formula/rule used to estimate the parameter | x̄ = (Σxᵢ) / n — the sample mean formula |
| Estimate (Statistic) | The specific value from one application of the estimator | x̄ = 103.4 (from a particular sample of 200 adults) |
| Sampling Distribution | The distribution of all possible values of the statistic across repeated samples | Distribution of all possible x̄ values for n = 200 |
Jerzy Neyman and Egon Pearson distinguished between estimators and estimates in the 1930s, formalizing the idea that we should evaluate the estimator's long-run behavior (is it unbiased? efficient? consistent?) rather than any single estimate. A good estimator is unbiased (E[x̄] = μ), efficient (minimum variance among unbiased estimators), and consistent (converges to the parameter as n → ∞). The sample mean satisfies all three properties for estimating μ, which is why it is used so universally.
The SAMPLE Framework: Identifying Parameters in Any Study
The SAMPLE framework gives a reproducible six-step process for identifying the statistic, parameter, population, and sample in any research scenario. This original framework is designed for students moving from descriptive to inferential statistics.
Specify the Population
Define exactly who or what the study is about. The population sets the scope: "all U.S. adults," "all patients with Type 2 diabetes," "all bolts produced in this factory run."
Acquire a Sample
Identify the subset that was actually measured. Note n (sample size) and the sampling method. Random sampling is required for statistics to be unbiased estimators of parameters.
Measure the Statistic
Compute the sample summary: x̄, s, p̂, or r from the sample data. Label it correctly with the Roman-letter notation. This is the known value from your data.
Predict the Parameter
State the corresponding Greek-letter parameter you are estimating: μ, σ, p, or ρ. Confirm it is unknown. The statistic is your best current estimate of it.
Learn from Inference
Apply the appropriate inferential method — confidence interval or hypothesis test — to draw a conclusion about the parameter based on the statistic.
Evaluate Uncertainty
Report the margin of error, confidence interval, or p-value. Uncertainty in estimating a parameter is unavoidable from sample data — quantifying it is what makes the inference trustworthy.
Decision Guide: Statistic or Parameter?
Three questions identify whether any given number is a statistic or a parameter.
Is This Number a Statistic or a Parameter?
When a study measures everyone in the population (a census), the resulting numbers are parameters — even though they are known. For example, if a company surveys all 80 of its employees and computes a mean satisfaction score, that mean is μ (a parameter), not x̄ (a statistic), because the entire population was measured. The U.S. Census Bureau uses this distinction explicitly in its statistical publications.
Real-World Applications
The statistic-parameter framework appears in every field that uses data to make decisions about populations larger than what can be directly measured.
Clinical Research
Sample statistics from trial participants (x̄ of n = 300 patients) estimate population parameters (μ for all patients with the condition). Regulatory agencies require confidence intervals around efficacy parameters.
Political Polling
Sample proportion p̂ from 1,000 surveyed voters estimates the population proportion p of all eligible voters. The margin of error reflects SE = √[p̂(1−p̂)/n].
Quality Control
Sample mean x̄ and sample standard deviation s from randomly inspected units estimate population parameters μ and σ for the entire production run. Control charts track whether x̄ stays within limits set by σ.
Finance & Economics
Sample return statistics (x̄, s) from historical data estimate population parameters for future expected return and volatility. Risk models like Value at Risk rely on parameter estimates from sample data.
Public Health
Disease prevalence is a population parameter p (true proportion infected). Surveys estimate it with p̂. Confidence intervals guide resource allocation and intervention policy across the full population.
Machine Learning
Model weights learned from training data are statistics (estimates). The "true" weights that would minimize loss over all possible data represent population parameters. Overfitting means fitting sample statistics too closely at the expense of parameter accuracy.
Software Examples: Excel, R, and Python
Statistical software distinguishes between sample statistics and population parameters through different functions. The table below shows exactly which function computes which quantity in the three most common tools.
| Quantity | Excel | R | Python (NumPy / SciPy) |
|---|---|---|---|
| Sample Mean (x̄) | =AVERAGE(data) | mean(data) | np.mean(data) |
| Sample SD (s) | =STDEV.S(data) | sd(data) | np.std(data, ddof=1) |
| Population SD (σ) | =STDEV.P(data) | — | np.std(data, ddof=0) |
| Sample Variance (s²) | =VAR.S(data) | var(data) | np.var(data, ddof=1) |
| Population Variance (σ²) | =VAR.P(data) | — | np.var(data, ddof=0) |
| Sample Proportion (p̂) | =COUNTIF(data,1)/COUNT(data) | mean(data == 1) | np.mean(data == 1) |
| Standard Error (SE) | =STDEV.S(data)/SQRT(COUNT(data)) | sd(data)/sqrt(length(data)) | scipy.stats.sem(data) |
The key distinction in Python is the ddof parameter (degrees of freedom correction). Setting ddof=1 computes the sample statistic (divides by n−1); ddof=0 computes the population parameter (divides by N). In R, sd() and var() always use the sample formula (n−1) by default. In Excel, the .S suffix means "sample" and the .P suffix means "population." This distinction directly mirrors the statistic-vs-parameter distinction discussed throughout this guide.
Sample Mean vs Population Mean Calculator
Enter your data below to compute both the sample mean (x̄) and, if you have the full population, the population mean (μ). The calculator also shows sample standard deviation (s), population standard deviation (σ), and standard error (SE).
Sample Statistics vs Population Parameters Calculator
Common Misconceptions
| Misconception | What Is Actually True |
|---|---|
| ✗ "A larger sample makes a statistic into a parameter." | ✓ No sample, regardless of size, is a parameter unless it covers the entire population. A large sample gives a more precise estimate of the parameter, but it remains a statistic. |
| ✗ "If the population is small, the summary is a statistic." | ✓ Population size is irrelevant. If you measure every member — even just 10 people — the resulting summary is a parameter. If you measure a subset, it is a statistic. |
| ✗ "Parameters and statistics are different types of numbers — one is more accurate." | ✓ They are the same types of numbers (means, proportions, etc.). The difference is where they come from: sample vs population. A well-designed sample produces a statistic that is a good approximation of the parameter. |
| ✗ "The population mean μ changes from study to study." | ✓ μ is fixed — it is the true value for a specific population at a specific time. What changes from sample to sample is the statistic x̄, which is why we build confidence intervals around x̄ rather than μ. |
| ✗ "Statistic and parameter are just two words for the same idea." | ✓ They represent fundamentally different concepts. A parameter is what you want to know; a statistic is what you can compute. Confusing them leads to incorrect interpretations of confidence intervals, p-values, and effect sizes. |
| ✗ "The sample SD (s) divides by n for the same reason σ does." | ✓ The sample SD divides by n−1 (Bessel's correction), not n. Dividing by n systematically underestimates σ² because the sample mean x̄ is always closer to the sample data than μ is, consuming one degree of freedom. |
Statistic vs Parameter: Reference Cheat Sheet
- Statistic: From a sample | Known | Variable | Roman letters (x̄, s, p̂, r) | Estimates the parameter
- Parameter: From a population | Unknown | Fixed | Greek letters (μ, σ, p, ρ) | The target of inference
- Sample mean: x̄ = Σxᵢ / n | estimates → μ = Σxᵢ / N
- Sample SD: s = √[Σ(xᵢ−x̄)²/(n−1)] | estimates → σ = √[Σ(xᵢ−μ)²/N]
- Sample proportion: p̂ = x/n | estimates → p = X/N
- Standard error: SE = s/√n — how precisely x̄ estimates μ
- Greek rule: μ, σ, σ², ρ = parameters. x̄, s, s², r = statistics. (Exception: p and p̂ for proportions)
- The hat rule: Any symbol with a hat (^) is a sample estimate: p̂ estimates p, μ̂ estimates μ
- Census: If the full population is measured, the result is a parameter, not a statistic
Entity and Formula Glossary
| Symbol | Name | Type | Definition |
|---|---|---|---|
| x̄ | Sample Mean | Statistic | Average of sample values: Σxᵢ / n |
| μ | Population Mean | Parameter | True average of all population values: Σxᵢ / N |
| s | Sample Standard Deviation | Statistic | Spread of sample data: √[Σ(xᵢ−x̄)² / (n−1)] |
| σ | Population Standard Deviation | Parameter | True spread in population: √[Σ(xᵢ−μ)² / N] |
| s² | Sample Variance | Statistic | Σ(xᵢ−x̄)² / (n−1) |
| σ² | Population Variance | Parameter | Σ(xᵢ−μ)² / N |
| p̂ | Sample Proportion | Statistic | Fraction with the attribute in the sample: x/n |
| p | Population Proportion | Parameter | True fraction with the attribute in the population |
| n | Sample Size | Statistic | Number of observations in the sample |
| N | Population Size | Parameter | Total number of individuals in the population |
| r | Sample Correlation | Statistic | Pearson correlation coefficient for sample data |
| ρ | Population Correlation | Parameter | True linear correlation in the population |
| SE | Standard Error | Derived | SE = s/√n — precision of x̄ as an estimate of μ |
| CI | Confidence Interval | Derived | Range of plausible parameter values: x̄ ± z*(SE) |