Direct answer
A useful dataset description combines a measure of center, at least one measure of spread, a view of shape, and an inspection of unusual observations; no single summary can preserve all four.
Visual explanation
One dataset, several descriptive questions
What this calculation tells you
Descriptive statistics compress entered observations into a small set of features. They can reveal a typical level, how much observations differ, whether the distribution is symmetric or skewed, and which values deserve investigation.
Begin by identifying the measurement unit, observation unit, population of interest, and whether the entered values are a sample or the complete population. Then choose summaries that preserve the features relevant to the decision.
Where it is used
Quality and manufacturing
Summarize measured output while retaining variation and unusual runs.
Research
Describe a sample before selecting an inferential model.
Business operations
Compare transaction, service, or delivery records without hiding dispersion.
Education
Introduce the complementary roles of numerical summaries and graphics.
When this guide helps
- Preparing a first profile of a newly collected dataset.
- Checking whether an average hides skew or extremes.
- Comparing two groups with similar centers.
- Deciding which inferential questions are defensible next.
Start with the statistical question
Begin by identifying the measurement unit, observation unit, population of interest, and whether the entered values are a sample or the complete population. Then choose summaries that preserve the features relevant to the decision.
The accompanying visual follows one dataset through a dot plot, center markers, an interval for the middle half, and a shape outline. It shows why the same mean can coexist with very different spread and asymmetry.
Worked example
For 4, 5, 5, 6, 10, the mean is 6 while the median is 5. The high observation raises the mean and widens the range, so reporting only ‘average 6’ hides the concentration around 5 and 6.
Assumptions that carry the result
Every observation must describe the same variable on a compatible scale. Collection methods, missingness, repeated measurements, weighting, censoring, and dependence remain part of the study design and cannot be repaired by a descriptive summary.
Interpret the result without overreaching
A pattern in the entered data does not establish causation, data quality, or a population conclusion. Preserve raw observations, document exclusions, and distinguish an exploratory signal from confirmatory evidence.
- Reporting a mean without spread or sample size.
- Treating a convenient dataset as representative.
- Deleting an unusual value before investigating it.
Worked dataset: one high value
Values are 2, 3, 3, 5 and 20.
n = 5, sum = 33, mean = 6.6, median = 3, mode = 3, minimum = 2, maximum = 20 and range = 18.
The mean lies well above the median because 20 pulls it upward.
The summary does not prove 20 is an error or identify a probability distribution.
Reproduce this worked caseOpen Descriptive Statistics Calculator
Worked case: remove a value only by rule
Suppose 20 is confirmed as a unit-entry error and corrected to 2.0.
Corrected mean = (2+3+3+5+2)/5 = 3; median = 3; range = 3.
Correction changes the summary because the underlying evidence changed.
Do not delete an inconvenient value without documented verification.
Reproduce this worked caseOpen Descriptive Statistics Calculator
dataset description: compare assumptions, not just answers
Start with raw values and provenance. Statistics summarize what was entered; they cannot decide data validity by themselves.
| Case | Calculation focus | Interpretation |
|---|---|---|
| Original | 2, 3, 3, 5, 20 | Right-skewed entered set |
| Verified correction | 2, 2, 3, 3, 5 | Different dataset |
dataset description: calculation checklist
- Count and missing values shown
- Units compatible
- Raw data retained
- Centre and spread together
- No causal claim
Practical questions
Frequently asked questions
Which statistic should I calculate first?
Start with sample size and a plot, then report center and spread that fit the scale and shape. There is no universal single first summary.
Can descriptive statistics prove a difference?
No. They describe the entered observations. Population inference requires a justified design and method.
Should I remove every outlier?
No. Investigate measurement, entry, sampling, and domain context before deciding whether any observation is invalid.
Further reading
Authoritative sources
Use these primary and professional resources to check definitions, conventions, or requirements that may extend beyond this guide.