Descriptive statistics summarise the data you have. Inferential statistics use that data to draw conclusions about a larger group you didn’t fully measure. Most real analyses use both: you describe first, then infer.
Same data, two kinds of question
A café asks 50 customers to rate their visit from 1 to 10. The mean is 7.2 and the standard deviation is 1.6.
- Descriptive: “The 50 customers we asked gave a mean rating of 7.2, and most ratings fell between about 5.6 and 8.8.” That is a plain fact about those 50 people. No uncertainty is involved.
- Inferential: “We are 95% confident the mean rating among all our customers is between 6.75 and 7.65.” Now we are talking about people we never asked, so the statement comes with a margin of error.
Run the café example in the confidence interval calculator
What falls under each
| Descriptive | Inferential |
|---|---|
| Mean, median, mode | Confidence intervals |
| Standard deviation, range, IQR | t-tests, ANOVA, chi-square tests |
| Percentages and frequency tables | p-values and significance |
| Histograms, box plots, bar charts | Regression used to predict or generalise |
| Correlation coefficient r as a summary | Testing whether a correlation is non-zero in the population |
Some numbers sit in both columns depending on how you use them. A correlation of 0.45 in your sample is descriptive. Asking whether the population correlation could be zero is inferential.
Why you should never skip the descriptive step
An inferential test can’t tell you that one of your values was typed as 750 instead of 7.5, or that half your data cluster at one end. A histogram or a five-number summary will. Look at your data before you test it. The descriptive statistics calculator gives you the summary numbers, a histogram and a box plot in one go.
When you don’t need inference at all
If you measured the whole population, descriptive statistics are the answer. A teacher with the scores of every student in her class doesn’t need a t-test to know whether the class average is above 70. It either is or it isn’t. Inference only comes in when you want to reach beyond the people or things you measured.
Check yourself
“62% of the 800 voters surveyed support the plan, ± 3.4%.” Which part is descriptive and which is inferential?
“62% of the 800 voters surveyed” is descriptive. The “± 3.4%” is inferential: it describes how far the true share among all voters might be from 62%.