Statistics workbench

Paste up to 50,000 finite decimal numbers. Small lists update directly; larger lists use Calculate or Ctrl/⌘+Enter. Spaces, line breaks and semicolons work in either number format. The 1.5 format additionally uses commas as separators; the 1,5 format uses a comma within a decimal number, so comma-separated CSV requires the 1.5 setting. No grouping separators, hexadecimal values or expressions. Exponent notation is allowed. Invalid tokens reject the whole calculation. Editing inputs removes old results and plots.

Variance and standard deviation use n−1 (sample estimates), and SEM is sample SD/√n. R7 quantiles interpolate at index (n−1)p. Geometric and harmonic means require strictly positive data. Shape measures are adjusted Fisher–Pearson skewness G1 and bias-adjusted excess kurtosis G2 (type 2); they require respectively at least three and four observations and nonzero dispersion. Dashes mean undefined values. All-unique samples say that no values repeat; repeated modal values show their frequency.

The box uses R7 Q 1/Q 3, median and mean; whiskers end at the most extreme observations within 1.5 IQR of the quartiles. Other points are potential outliers, not automatically errors. This differs from software using Tukey hinges. At most 200 distinct outlier positions are shown when needed, while the count includes every observation. Histogram bins are equal width; the final bin includes the maximum. Hover bars/marks for values. Both plots use vector geometry and normalized coordinates, including constant and very wide samples.

Binary64 inputs are accumulated exactly for sum and rounded mean, then other statistics use scaled floating-point arithmetic. Genuine overflow and selected underflow are marked separately from undefined results. There is no claim of arbitrary decimal precision. The 50,000-value and one-million-character limits keep browser work bounded. Input drafts, including large lists and unfinished values, remain in local browser storage through navigation and reload. Results are recomputed for small lists; large lists await Calculate. CSV exports summary statistics with full available numerical precision and explicit numerical-limit markers.