A percentile is not a percentage of a value
The 90th percentile is not 90% of the sum or 90% of the maximum. It locates a position in sorted data. With small discrete datasets, the interpolated result need not be one of the observations.
POSITION / TRANSPARENT METHOD
Percentiles
Choose a percentage position and locate it within your sorted observations.
This tab’s workspace
Interactive examples
Predefined synthetic data only. The link shares the example, never your numbers or metadata. Loading an example replaces current fields.
01Enter your data
Example⇧ Import or paste a table
Choose options explicitly. Missing cells are flagged, not skipped.
Download an example CSV ↓
Review the table before importing
Uses the columns and decimal format selected above. Row numbers refer to the parsed table including the header; completely blank rows may already have been ignored. No correction or exclusion is automatic.
At most 80 notices are displayed; every row is checked.
The report retains the original table, edits and excluded rows. It contains data: share deliberately. This is an editable local trail, not a certification.
Compare table columns
Select up to 8 numeric columns. Every selected cell must be valid; no individual missing values are removed. This is a descriptive comparison, not a test between groups.
Resume an analysis from a file
Open a .json file exported by StatNivo. Replaces current data and recalculates in this calculator; nothing is uploaded.
Raw data: up to 10,000 values per series.
Temporary storage is off unless you select the option.
02Read your result
Loading the local calculator…
Export, save and share
Undefined values: empty cells. No additional rounding in CSV files. CSV headings use stable English identifiers for software compatibility.
Temporary storage is off unless you select the option. Optional sessionStorage, off by default. Keeps text, table and settings across tools and languages in this tab. Browser session restoration can retain it. Use “Delete saved data” on a shared device. Nothing is uploaded.
LOOK BEYOND THE MEAN
The shape of your data
Scroll the chart horizontally to read the full scale. Frequencies are also available in the table.
Read frequencies as a table
How we get the result
Inspect observations and deviations
50 rows per page. Original row positions are retained; CSV includes all observations.
UNDERSTAND, DON’T JUST CALCULATE
The method in practice
h = (n − 1) · p; interpolate between adjacent values
A step-by-step example
For 10, 20, 30, 40 and p = 0.9, the zero-based position is (4 − 1) × 0.9 = 2.7. Interpolate 70% of the way from 30 to 40 to get 37. Percentile 0 is the minimum and percentile 100 is the maximum.
Comparing different tools
We use type 7 for both quartiles and percentiles. Other methods, such as nearest rank, may disagree without either program being broken. Report the method with the result when preparing a document.
Practical checks before using a result
Sample or population?
For 2, 4, 4, 4, 5, 5, 7, 9 the population SD is 2; sample SD is √(32/7) ≈ 2.13809. Choose the population formula when describing the full population of interest, not because the number is smaller.
Standard deviation or standard error?
Standard deviation describes the spread of observations. The usual standard error of the mean is sample SD / √n under independent sampling with finite variance. For sample SD 10 and n = 25, SEM = 2. Repeated dependent readings do not automatically count as 25 independent measurements.
Importing from Excel without shifting pairs
Paste the entire table, select its real delimiter and header, then preview it. In decimal-comma data, use a semicolon or tab as the column delimiter. If Y is missing in row 3, fix that row rather than moving row 4 up.