A statistic is a number that summarizes or describes something about a sample, which is a smaller group drawn from a larger population. If you survey 500 people in a city of 100,000 and calculate their average income, that average is a statistic. If you could somehow measure every single person in the city, that average would instead be called a parameter. Statistics (the field) is built on this core idea: using what you can measure from a smaller group to learn about a larger one you can’t fully observe.
Statistics vs. Parameters
The distinction between a statistic and a parameter is one of the most fundamental concepts in the field. A parameter describes an entire population. A statistic describes a sample pulled from that population. In practice, you almost never know the true parameter because measuring every member of a population is rarely possible. Instead, you collect a sample, compute a statistic, and use it to estimate the parameter.
For example, the true average height of all adults in a country is a population parameter. You’d need to measure every adult to know it exactly. But if you measure 2,000 randomly selected adults and calculate their average height, that number is a sample statistic. Statisticians use Greek letters for parameters (μ for the population mean, for instance) and standard letters for statistics (x̄ for the sample mean). This notation exists precisely because keeping the two concepts separate matters so much.
The Two Main Branches
The field of statistics splits into two broad categories: descriptive and inferential. They serve very different purposes.
Descriptive statistics summarize what you already have. They take a pile of raw numbers and compress them into something meaningful. This includes measures of central tendency (mean, median, mode), which identify the middle or typical value, and measures of spread (range, variance, standard deviation), which tell you how spread out the data is. A company reporting that its employees earn a median salary of $58,000 with a standard deviation of $12,000 is using descriptive statistics. The numbers don’t predict anything or test any theory. They just describe the data at hand.
Inferential statistics go further. They use sample data to draw conclusions about a larger population, make predictions, or test whether a pattern is real or just noise. Techniques like hypothesis testing, correlation analysis, and regression analysis all fall here. When a medical study concludes that a drug lowers blood pressure, that conclusion comes from inferential statistics applied to a sample of patients, not from testing every person on Earth.
Measures of Central Tendency
The mean, median, and mode are the three most common ways to identify a “typical” value in a data set, and each works differently. The mean is the classic average: add up all the values and divide by how many there are. It’s the most widely used, but it’s sensitive to extreme values. A single billionaire moving into a small town can dramatically shift the mean income upward even though nobody else got richer.
The median is the middle value when all data points are arranged in order. It’s more resistant to outliers, which is why housing prices and income data are often reported as medians rather than means. The mode is simply the most frequently occurring value. It’s the only measure of central tendency that works for non-numerical categories, like identifying the most popular car color in a parking lot.
Measures of Spread
Knowing the center of your data is only half the picture. Two data sets can have the same mean but look completely different. Imagine two classrooms where the average test score is 75. In one, every student scored between 70 and 80. In the other, scores ranged from 30 to 100. The averages are identical, but the distributions are not.
The range (highest value minus lowest) is the simplest measure of spread. Variance calculates how far each data point sits from the mean, on average, by squaring those distances. Standard deviation is the square root of variance, which brings the number back into the original units and makes it easier to interpret. A small standard deviation means data points cluster tightly around the mean. A large one means they’re scattered widely.
How Sampling Works
The reliability of any statistic depends heavily on how the sample was collected. A poorly chosen sample produces misleading numbers no matter how sophisticated the analysis. Several standard sampling methods exist, each with trade-offs.
- Random sampling gives every member of the population an equal chance of being selected. It’s the gold standard for avoiding bias.
- Systematic sampling selects every kth person from a list, like surveying every 10th customer who walks through a door.
- Stratified sampling divides the population into subgroups by a characteristic (age, gender, income level) and then samples from each subgroup. This ensures that important categories are represented.
- Cluster sampling divides the population into geographic or organizational groups, randomly selects some of those groups, and then surveys everyone within them.
Statistical Significance and P-Values
When researchers test whether a result is real or just a fluke, they typically calculate a p-value. This is the probability of seeing a result as extreme as the one observed if nothing interesting were actually going on (if the “null hypothesis” were true). A small p-value suggests the result is unlikely to be random chance.
The conventional threshold is 0.05, meaning there’s a 5% or lower probability the result occurred by chance alone. This cutoff traces back to the statistician Ronald Fisher, who in 1925 described it as a “convenient” dividing line. It corresponds roughly to being more than two standard deviations from the mean of a normal distribution. Fisher never intended it as a rigid rule, and even at the time, other statisticians called it “quite arbitrary.” Yet it became the standard that researchers still use today.
A confidence interval is a related tool. Instead of giving a single yes-or-no answer about significance, it provides a range of values within which the true population parameter likely falls. A 95% confidence interval, for instance, means that if you repeated the study many times, about 95% of the resulting intervals would contain the true value. P-values and confidence intervals are mathematically connected: if a confidence interval doesn’t include the “no effect” value, the corresponding p-value will be below 0.05.
Correlation Is Not Causation
One of the most common errors in interpreting statistics is assuming that because two things move together, one must cause the other. Correlation measures the strength and direction of a relationship between two variables, expressed as a value between -1 and 1. But a strong correlation alone says nothing about whether one variable actually influences the other.
Smoking is correlated with alcoholism, for example, but smoking doesn’t cause alcoholism. Both may be driven by shared underlying factors. Ice cream sales and drowning rates both rise in summer, but buying ice cream doesn’t put anyone at risk of drowning. The heat drives both. This distinction matters enormously in fields like medicine and public policy, where confusing correlation with causation can lead to ineffective or harmful decisions.
Statistics in Modern Technology
Many of the tools people interact with daily, from search engines to voice assistants, run on statistical methods. Machine learning, the technology behind most modern artificial intelligence, is fundamentally built on statistics. Regression, one of the oldest statistical techniques, powers predictions in applications ranging from forecasting blood pressure to estimating life expectancy. Neural networks, which drive image recognition and language processing, use a process of adjusting weights and thresholds to minimize prediction errors, a goal expressed mathematically through the same sum-of-squared-errors formula that statisticians have used for over a century.
Classification algorithms like the Naïve Bayes classifier apply Bayes’ theorem, a foundational concept in probability, to sort emails into spam and not-spam or flag fraudulent transactions. The core logic is identical to what a statistics textbook teaches: given what you know about prior probabilities, update your beliefs when new evidence arrives. The scale is different, but the principles are the same ones that define a statistic in the first place: summarize data, find patterns, and use what you observe to make informed guesses about what you haven’t observed.