What Is a Population in Statistics?
The word "population" in statistics does not always mean people. A population can be a set of factory-produced bolts, all blood pressure readings taken under a specific protocol, every webpage indexed by a search engine, or every possible random number a die could produce. The defining feature is that the set is completely specified by the research question before data collection begins.
When you can measure every member of the population, you compute a population parameter — a fixed, exact number. The average annual salary of all 500 employees at one company, for example, is a parameter. You have the payroll. You can compute it exactly. Parameters use Greek letters in standard statistical notation: μ for the population mean, σ for population standard deviation, and σ² for population variance.
What Is a Sample in Statistics?
When you compute a numerical summary from a sample, you get a sample statistic. The average salary of 50 randomly chosen employees is a statistic. Because you didn't measure everyone, the statistic will differ slightly from the true population parameter. That difference is sampling error, and it is expected — it does not mean the study was done incorrectly.
Sample statistics use Roman letters: x̄ (x-bar) for the sample mean, s for sample standard deviation, and s² for sample variance. The relationship between statistics and parameters is the foundation of hypothesis testing and confidence intervals — both of which use sample statistics to draw conclusions about population parameters.
A population is the entire group a researcher wants to study, described by parameters (μ, σ). A sample is a smaller subset drawn from that population, described by statistics (x̄, s). Researchers study samples because measuring every member of a population is usually impractical, too costly, or impossible. A well-drawn random sample allows researchers to estimate population parameters accurately using inferential statistics.
Population vs Sample: Key Differences
The table below compares population and sample across every dimension that appears in an introductory statistics course — from notation and formulas to cost, feasibility, and the type of analysis each supports.
| Feature | Population | Sample |
|---|---|---|
| Definition | All members of the group of interest | A selected subset of the population |
| Size notation | N (capital) | n (lowercase) |
| Mean notation | μ (mu) | x̄ (x-bar) |
| Std. deviation notation | σ (sigma) | s |
| Variance notation | σ² | s² |
| Proportion notation | P | p̂ (p-hat) |
| Numerical summary type | Parameter (exact) | Statistic (estimate) |
| Standard deviation formula divisor | N | n − 1 (Bessel's correction) |
| Feasibility | Often impractical or impossible | Practical and commonly used |
| Cost | Very high (census) | Lower |
| Time required | Long | Shorter |
| Accuracy | Exact (if measured completely) | Approximate (subject to sampling error) |
| Statistical use | Descriptive statistics only | Inferential statistics (hypothesis tests, CIs) |
| Example | All 40,000 students at a university | 400 randomly selected students |
Population Parameter vs Sample Statistic: Notation Reference
The single most tested distinction in introductory statistics is the notation difference between population parameters and sample statistics. Parameters are always represented by Greek letters; statistics by Roman letters.
| Measure | Population Parameter | Sample Statistic | Description |
|---|---|---|---|
| Mean | μ (mu) | x̄ (x-bar) | Average value |
| Standard deviation | σ (sigma) | s | Spread around the mean |
| Variance | σ² | s² | Squared spread (average squared deviation) |
| Proportion | P (or π) | p̂ (p-hat) | Fraction with a characteristic |
| Correlation | ρ (rho) | r | Linear relationship strength |
| Size | N | n | Number of members |
| Regression slope | β (beta) | b | Rate of change in regression |
Key Formulas: Population vs Sample
The mean formula is the same whether you're computing a population parameter or a sample statistic — you sum all values and divide by the count. The standard deviation formula differs: for a sample, you divide by n − 1 rather than n. This adjustment is called Bessel's correction, and it corrects for the fact that a sample tends to underestimate the true spread of the population.
Population Mean vs Sample Mean
Σxᵢ = sum of all values
N = population size
Σxᵢ = sum of sample values
n = sample size
Population Standard Deviation vs Sample Standard Deviation
μ = population mean
N = population size
x̄ = sample mean
n − 1 = degrees of freedom
Dividing by n underestimates how spread out the population really is, because the sample values naturally cluster around the sample mean rather than the population mean. Dividing by n − 1 corrects for this bias, making s² an unbiased estimator of σ². This is tested on almost every introductory statistics exam. For more, see the guide to standard deviation.
Census vs Sample Survey
A census is the special case where you measure every member of the population — no sampling involved. The U.S. Census Bureau conducts a national census every 10 years to count every person living in the United States. This gives an exact population count with no sampling error. But it costs billions of dollars and takes years to process.
For everything else — estimating poverty rates, tracking disease prevalence, measuring unemployment — the Census Bureau uses sample surveys like the American Community Survey, which collects data from roughly 3.5 million households each year instead of all 130 million. This gives near-census accuracy at a small fraction of the cost.
| Feature | Census | Sample Survey |
|---|---|---|
| Who is measured | Every member of the population | A selected subset |
| Sampling error | None (complete data) | Present (by design) |
| Cost | Very high | Significantly lower |
| Time | Long | Faster |
| Result type | Parameter (exact) | Statistic (estimate) |
| Feasibility for large groups | Difficult or impossible | Practical |
| U.S. example | Decennial Census (every 10 years) | American Community Survey (ongoing) |
| Clinical example | Recording every patient's blood type at a hospital | Testing a random sample in a clinical trial |
Decision Guide: Should You Use a Census or a Sample?
Population vs Sample Examples Across 6 Fields
The distinction between population and sample depends on what question you are asking, not on the size of the group. A population of 50 people is still a population if you measure all 50. A group of 10,000 is a sample if you only selected them from a larger universe.
Healthcare Research
Scenario: A hospital wants to estimate the average fasting blood glucose level of its 8,000 registered diabetic patients.
Population: All 8,000 registered diabetic patients at the hospital (N = 8,000). The parameter of interest is μ, the true mean fasting blood glucose.
Sample: 200 patients selected using simple random sampling (n = 200). Their blood glucose readings are collected and averaged to get x̄.
Result: x̄ = 142 mg/dL. This sample statistic estimates μ for the full population. A 95% confidence interval then quantifies the uncertainty around that estimate.
✅ The 200-patient sample is a subset of the 8,000-patient population. The sample mean (x̄ = 142 mg/dL) estimates the population mean (μ), which cannot be computed without testing all 8,000 patients.
Education Research
Scenario: A school district wants to evaluate the reading level of its 12,000 third-grade students before designing a new curriculum.
Population: All 12,000 third-grade students in the district. The parameter σ represents the true spread in reading scores across all students.
Sample: 500 students drawn using stratified sampling — students selected proportionally from each school in the district (n = 500).
Result: The sample mean reading score is x̄ = 74.3 and the sample standard deviation is s = 11.6. The district uses these to estimate μ and σ for the full population and plan curriculum resources accordingly.
✅ Stratified sampling ensures every school's students are represented. The sample statistics x̄ and s are used to estimate the population parameters μ and σ without testing all 12,000 students.
Business and Market Research
Scenario: An e-commerce company wants to know what proportion of its 500,000 active customers prefer paying by digital wallet over credit card.
Population: All 500,000 active customers (N = 500,000). The population proportion P is the true fraction who prefer digital wallet.
Sample: 1,000 customers are randomly selected and surveyed (n = 1,000). The sample proportion p̂ is calculated.
Result: 430 of 1,000 say digital wallet. p̂ = 430/1,000 = 0.43. The company estimates that about 43% of all 500,000 customers prefer digital wallet, with a margin of error calculated from the confidence interval for a proportion.
✅ The sample proportion (p̂ = 0.43) estimates the unknown population proportion (P). Surveying all 500,000 customers would be prohibitively expensive for a simple preference question.
Clinical Trials
Population: all patients with a condition worldwide. Sample: enrolled trial participants. Regulatory bodies like the FDA require well-defined sampling protocols before approving treatments.
Election Polling
Population: all eligible voters. Sample: 800–1,200 likely voters surveyed. The margin of error reported in polls directly reflects sampling error from studying a sample rather than the full electorate.
Quality Control
Population: all units produced on a manufacturing line. Sample: a batch inspected each hour. Testing every unit would halt production; statistical sampling maintains quality with minimal disruption.
A/B Testing
Population: all future website visitors. Sample: users randomly assigned to version A or B during the test window. Results are used to infer which version performs better for the full user base.
Sampling Error and Sampling Bias
Two concepts are often confused but mean very different things for research quality.
What Is Sampling Error?
If the true average exam score across all students is μ = 75, and a random sample of 30 students yields x̄ = 73.2, the sampling error is −1.8 points. A different random sample of 30 would likely give a different x̄ — some above 75, some below. The distribution of all possible sample means is the sampling distribution, and its spread is measured by the standard error:
s = sample standard deviation
n = sample size
SE decreases as n increases
Larger samples produce smaller standard errors, which means sample statistics cluster more tightly around the true population parameter. This is the mathematical basis for why bigger samples give more accurate estimates — a consequence of the Central Limit Theorem and the Law of Large Numbers.
What Is Sampling Bias?
Sampling error is random and decreases with larger samples. Sampling bias is systematic — it pulls estimates in one direction regardless of sample size. A biased sampling method cannot be fixed by collecting more data. The source must be identified and eliminated through better study design. For related coverage see the guide to study design.
Sampling bias occurs when some members of the population have a different probability of being selected than others, and that difference is related to the outcome being studied. Classic examples include voluntary response bias (only people with strong opinions respond to a survey), convenience sampling bias (only surveying people who are easy to reach), and undercoverage (systematically missing certain subgroups).
| Bias Type | What Goes Wrong | Real-World Example |
|---|---|---|
| Voluntary response | Only motivated respondents reply | Online polls where people choose to participate |
| Convenience sampling | Only accessible people are included | Surveying shoppers at one mall to represent all consumers |
| Undercoverage | Some groups are systematically excluded | Phone surveys missing people without landlines |
| Non-response bias | Those who don't respond differ from those who do | Customer satisfaction surveys returned mostly by unhappy customers |
| Leading questions | Question wording pushes toward certain answers | "Don't you agree that taxes are too high?" surveys |
5 Types of Sampling Methods
How a sample is selected determines whether it is likely to be representative. The five methods below cover everything from fully random approaches to practical non-probability designs. For a deeper treatment of each, see the complete guide to study design and sampling on Statistics Fundamentals.
Simple Random Sampling
Every member of the population has an equal chance of selection. Example: drawing 50 employee IDs from a hat. This is the gold standard for unbiased estimates and the method assumed in most hypothesis tests.
Stratified Sampling
The population is divided into subgroups (strata) and samples are drawn from each stratum. Example: sampling students proportionally from each grade level. Guarantees representation of all subgroups.
Cluster Sampling
The population is divided into clusters (e.g., city blocks), a random set of clusters is chosen, and all members within those clusters are measured. More practical for geographically dispersed populations.
Systematic Sampling
Every kth member is selected after a random start. Example: every 10th person on a voter registration list. Simple to implement; can introduce bias if the list has a periodic pattern.
Convenience Sampling
Selecting whoever is easiest to reach. Useful for pilot studies and exploratory research. Results cannot be generalized to the full population because the sample is unlikely to be representative.
Interactive: Population or Sample Identifier
Enter details about any research scenario and the tool will classify whether each group described is a population or a sample, and identify the likely parameter or statistic of interest.
Population vs Sample Identifier
How Large Does a Sample Need to Be?
Sample size is one of the most practical questions in applied statistics. Too small and your estimates are unreliable; too large and you waste resources. The right answer depends on how much accuracy you need, how variable the population is, and what level of confidence you require.
z* = critical value (1.96 for 95% CI)
σ = population standard deviation
E = desired margin of error
A researcher wants to estimate average daily screen time in adults with a margin of error of ±15 minutes at 95% confidence. Prior studies suggest σ ≈ 60 minutes. What sample size is needed?
Identify values: z* = 1.96 (95% CI), σ = 60, E = 15
Apply formula: n = (1.96 × 60 / 15)² = (117.6 / 15)² = (7.84)² = 61.5 → round up to 62
Interpret: A sample of at least 62 adults will produce a margin of error ≤ 15 minutes at 95% confidence.
✅ n = 62. Verify this with the full sample size calculator.
The SAMPLE Framework for Research Planning
Before collecting any data, researchers need to define both their population and their sampling strategy. The SAMPLE framework below provides a structured six-step process for moving from a research question to a valid sample design.
The SAMPLE Framework
Common Misconceptions About Population and Sample
| Misconception | ❌ Wrong Thinking | ✅ Correct Understanding |
|---|---|---|
| Size determines which it is | "The group has 50,000 people, so it must be a sample." | A group of 50,000 is a population if it's the entire group you're studying. A group of 10 is a sample if it was selected from a larger group. |
| Larger sample = no sampling error | "If my sample is big enough, the sample mean will exactly equal μ." | Sampling error approaches zero as n → ∞, but it never equals zero for a finite sample. Larger samples reduce uncertainty; they don't eliminate it. |
| Random = representative | "I picked randomly, so my sample must represent the population." | Random sampling makes a sample likely to be representative on average, but any single random sample may not perfectly mirror the population. That's what confidence intervals account for. |
| σ and s are the same formula | "The standard deviation formula is always Σ(xᵢ − mean)² / n." | Population SD divides by N. Sample SD divides by n − 1 (Bessel's correction). Using the wrong formula in a hypothesis test produces incorrect results. |
| Sampling bias and sampling error are the same | "Any error in my sample result is sampling error." | Sampling error is random variation from taking a subset. Sampling bias is a systematic, directional distortion from a flawed selection process. More data fixes error; it does not fix bias. |
Original Dataset: Student Exam Scores (Population vs Sample)
The table below contains an original dataset of exam scores for 20 students from a fictional class of 100. The first column shows population data (all 20 in a small section); the second shows a random sample of 8 drawn from those 20. Both parameter and statistic values are computed from this data.
| Student ID | Score (Population, N=20) | In Sample (n=8)? |
|---|---|---|
| S01 | 88 | ✓ Yes |
| S02 | 74 | — |
| S03 | 91 | ✓ Yes |
| S04 | 65 | — |
| S05 | 78 | ✓ Yes |
| S06 | 82 | — |
| S07 | 55 | ✓ Yes |
| S08 | 93 | — |
| S09 | 67 | ✓ Yes |
| S10 | 76 | — |
| S11 | 84 | ✓ Yes |
| S12 | 72 | — |
| S13 | 89 | ✓ Yes |
| S14 | 61 | — |
| S15 | 95 | ✓ Yes |
| S16 | 70 | — |
| S17 | 83 | — |
| S18 | 58 | — |
| S19 | 77 | — |
| S20 | 81 | — |
This dataset shows that the sample mean (x̄ = 80.6) overestimates the true population mean (μ = 77.9) by 2.7 points — that gap is the sampling error for this particular sample. A different set of 8 students would have produced a different x̄, with a different sampling error, illustrating why repeated sampling produces a distribution of x̄ values centered around μ. This is the core idea behind the Central Limit Theorem.
Glossary: Population vs Sample Key Terms
| Term | Symbol | Definition |
|---|---|---|
| Population | — | The complete set of all individuals or measurements of interest in a study |
| Sample | — | A subset of the population selected to represent it |
| Population Size | N | The number of members in the population |
| Sample Size | n | The number of members in the sample |
| Population Mean | μ | The average of all population values |
| Sample Mean | x̄ | The average of all sample values; estimates μ |
| Population Standard Deviation | σ | The spread of the entire population; formula uses N |
| Sample Standard Deviation | s | The spread of the sample; formula uses n − 1 |
| Population Variance | σ² | Square of the population standard deviation |
| Sample Variance | s² | Square of the sample standard deviation; unbiased estimator of σ² |
| Population Proportion | P or π | The true fraction of the population with a characteristic |
| Sample Proportion | p̂ | The observed fraction in the sample; estimates P |
| Parameter | Greek letter | A numerical summary of the population; fixed but often unknown |
| Statistic | Roman letter | A numerical summary of a sample; used to estimate a parameter |
| Sampling Error | x̄ − μ | The difference between a sample statistic and the population parameter |
| Standard Error | SE = s/√n | The standard deviation of the sampling distribution of x̄ |
| Sampling Bias | — | Systematic distortion from a flawed selection process; not fixed by more data |
| Census | — | Measuring every member of the population; produces exact parameters |
| Sampling Frame | — | The list of population members from which the sample is drawn |
| Inferential Statistics | — | Using sample statistics to draw conclusions about population parameters |
Frequently Asked Questions
What to Study Next
Understanding population versus sample is the prerequisite for almost every topic in inferential statistics. The concepts here appear directly in the following areas, each covered in depth on Statistics Fundamentals:
Sampling Distributions
How sample means are distributed around μ. The foundation for understanding why x̄ reliably estimates μ across repeated samples.
Confidence Intervals
How to construct a range of plausible values for μ based on x̄, s, and n. Every CI formula directly uses the population/sample distinction.
Hypothesis Testing
Using sample statistics to test claims about population parameters. The z-test and t-test are built on the population vs sample framework.
Descriptive Statistics
Computing and interpreting mean, median, mode, variance, and standard deviation — for both populations and samples.
Normal Distribution
The sampling distribution of x̄ is approximately normal for large n (by CLT). The normal distribution governs z-scores and critical values.
Sample Size Calculator
Compute the minimum n needed for a desired margin of error and confidence level, for both means and proportions.