KURT Function (Gnumeric)

• Introduced: Gnumeric 0.19 release notes record its addition.
• Supported in: Gnumeric 1.12.62; no retirement is established.
• Scope: Desktop Gnumeric with its Statistics plugin. English comma-separated formulas are shown.
• Checked: October 14, 2026. The tables and named advanced check use 1.12.57 ssconvert. Current 1.12.62 source was inspected; that release was not executed.

KURT calculates bias-corrected sample excess kurtosis. The subtraction of 3 gives the normal-distribution reference value zero. It is a measure of standardized fourth-moment shape, not a percentage, probability, or automatic outlier test. A finite sample from a normal population need not return exactly zero. Versioned function help.

=KURT(value1[,value2,...])

• value1,...: numbers or ranges containing observations. At least four usable numbers and nonzero variance are needed.
• Referenced text, logical values and empty cells are excluded. Direct text and logical arguments need the version distinction below.

Brackets mark optional arguments; do not type them. Current implementation.

Enter these columns as numbers, except for the explicitly typed H column. F7 and H7 are genuinely empty.

Use an apostrophe entry marker if needed to store H5 as text; it is not part of the stored value. The results below are native outputs rounded to 12 decimal places for readability, rather than application screenshots.

For five recorded processing times 1,2,3,4,10, enter the numbers in F2:F6. =KURT(F2:F6) returns approximately 3.152. Retain the 10 unless there is an independent reason to correct or exclude it; a large contribution to kurtosis is not evidence of a data-entry mistake. Report the five-value sample and its size alongside the statistic.

=KURT(N2:N5) gives −1.2 for 1,2,3,4. This is a different shape from the five-time sample, not a negative amount of variability.

Let n be the numeric count, m the mean, and s² = Σ(x−m)²/(n−1). Gnumeric computes n(n+1)Σ((x−m)/s)^4 / ((n−1)(n−2)(n−3)) − 3(n−1)²/((n−2)(n−3)).

For F2:F6, the population excess kurtosis is −0.212. The correction (n−1)((n+1)g₂+6)/((n−2)(n−3)) gives 3.152 when n=5 and g₂=−0.212. Thus sample KURT and population KURTP are deliberately different. The native check =KURT(F2:F6)-KURT(P2:P6) gives zero: multiplying every observation by 2 does not change this shape measure. Neither do a common shift or reflection, apart from floating-point roundoff.

The 1.12.57 callback ignores direct strings and logical values. The 1.12.62 callback adds direct-string parsing: numeric text becomes a number and failed numeric conversion gives #VALUE!. Only the sample variants KURT and SKEW also add direct-boolean conversion. Referenced strings and booleans remain excluded. See the current collector.

The exact first release of these collection changes is not established. The comparison concerns the stated English inputs; locale-dependent numeric text was not tested.

Fewer than four retained numeric values, an empty set, or a constant set produces #DIV/0!. An existing cell error is propagated. Check the numeric count after filtering; a five-cell range can contain only one numeric observation.

• Current worksheet implementation
• Tested-version worksheet implementation
• Current moment calculation

Both the stale SUM cache and the uncached SUM recalculated correctly. Saved formula identities and typed inputs were checked in the native workbook exports. The importer reported unsupported style-extension elements; this does not establish complete formatting support. GUI entry, other locales and later-version execution were not tested.

AVERAGE, VAR, STDEV.