Choose plain-list input for up to 10,000 observations: commas, spaces, semicolons and newlines separate values. Choose grouped-number input explicitly for thousands commas; then separate observations with semicolons or newlines. A malformed token blocks the sample.
Read the result
R-7 is the default linear estimator; R-6 and R-8 use different positions with endpoint extension, while R-1 uses ceiling rank. An estimated quantile may lie between observations and need not have exactly p% of this sample at or below it.
For L observations below x and E equal to x, strict-below is 100L/n, the empirical CDF is 100(L+E)/n, and midrank is 100(L+E/2)/n. These are distinct tie conventions. Sample SD uses n−1, with a displayed singleton convention of zero.
Conventions and limits
Fractional percentiles retain their precision in generated Excel, R and NumPy code. R defaults to type 7. R-6 code explicitly extends endpoints to match this calculator, since native Excel EXC has a narrower domain. Blank or invalid inputs clear current snippets and charts.
Compare methods and samples consistently
The single-quantile, batch-method and quantile-and-rank workflows all use equally weighted observations. Enter percentile requests on the 0–100 scale. For comparisons, keep the dataset and requested percentiles fixed while changing the estimator; the rank convention is a separate choice.
Keep duplicate observations, since removing ties changes the empirical distribution. Sorting changes order, not multiplicity. A saved scenario retains the data, method, parser convention and other active settings. Display rounding does not change the underlying result, and grouped-number parsing does not introduce statistical weights.