Direct comparison
Descriptive vs inferential statistics
Descriptive statistics summarise and describe the data actually collected; inferential statistics use that sample data to draw probabilistic conclusions about a larger population.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
Side-by-side comparison
| Dimension | Descriptive statistics | Inferential statistics |
|---|---|---|
| Purpose | Summarise and describe the data that were actually collected. | Use sample data to draw conclusions about a wider population. |
| Scope of claim | Says nothing beyond the dataset in hand. | Generalises from the sample to the population it represents. |
| Data used | The full set of data available, whether a sample or a census. | A sample drawn from the larger population of interest. |
| Typical measures | Mean, median, mode, range, variance, standard deviation, percentages. | Confidence intervals, hypothesis tests, p-values, regression estimates. |
| Role of probability | None required — it simply reports what the data show. | Central — conclusions are stated with quantified uncertainty. |
| Sampling needed | Not essential; describes whatever data exist. | Relies on sound, ideally random, sampling to be valid. |
| Output | Summary numbers, tables and charts. | Estimates, intervals and test decisions about parameters. |
| Key risk | Misleading summaries (for example, a mean distorted by outliers). | Sampling error and biased samples leading to wrong generalisations. |
| Example | The average exam mark in this class was 68%. | Estimating the mean mark of all students nationally, with a margin of error. |
Two stages of the same analysis
Descriptive and inferential statistics are not rivals but successive stages of analysis. Researchers almost always begin descriptively — computing summaries and plotting the data — to understand what was collected and to check it for errors and unusual values. Inferential methods then build on that foundation, using the sample to reason about the population it was drawn from. Inference depends on the central limit theorem and on sound sampling: if the sample is biased or too small, no inferential technique can rescue the conclusion. The two approaches therefore work together, with good description being the groundwork for trustworthy inference.
Common questions
FAQ
When should I use descriptive rather than inferential statistics?+
Use descriptive statistics when your aim is simply to summarise and present the data you have — the centre, spread and shape of a dataset, with means, standard deviations and charts. Use inferential statistics when you want to generalise from a sample to a wider population, estimate an unknown parameter, or test a hypothesis. Most analyses use descriptive statistics first, then inferential methods.
Is the mean a descriptive or an inferential statistic?+
The mean is a descriptive statistic: it summarises the central tendency of the data you have. It becomes part of inferential work when a sample mean is used to estimate an unknown population mean, or compared between groups in a hypothesis test. The same summary value can serve both roles depending on whether you are describing or generalising.
Can you have inferential statistics without descriptive statistics?+
In practice, no. Inferential methods are built on summaries such as means, proportions and variances computed from the sample — these are descriptive statistics. Sound description also lets you check assumptions and spot data problems before inference. Skipping the descriptive stage risks drawing population conclusions from data you have not properly understood.
Going deeper
