KURT in Collabora Office: Calculate Adjusted Excess Kurtosis
Use Collabora KURT to distinguish adjusted excess kurtosis from an unadjusted fourth moment and verify its sample-size requirements.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
Quick answer: for 1, 2, 3, 4 and 5 in A2:A6, =KURT(A2:A6) has independently expected result -1.2. This is an adjusted excess-kurtosis statistic, not a raw fourth moment.
Syntax and data requirements
=KURT(Argument1 [; Argument2 ...])
The dedicated Collabora help permits up to 255 arguments, including references and ranges, and requires at least four numeric values. Ordinary text and empty cells inside a referenced range are ignored. The number of usable observations can therefore be smaller than the number of selected cells.
Current source uses the sample standard deviation and a finite-sample adjustment, then subtracts an excess-kurtosis correction. A normal-model reference has excess kurtosis zero, whereas the corresponding non-excess kurtosis convention has reference three.
Do not compare values from different software or reports until their definitions match. A statistic called kurtosis can use different bias corrections or omit the subtraction of three.
Example 1: calculate an auditable symmetric dataset
Enter 1, 2, 3, 4 and 5 in A2:A6. The mean is 3 and sample variance 2.5. Centered values are -2, -1, 0, 1 and 2.
Their fourth powers are 16, 1, 0, 1 and 16, totaling 34. Dividing this sum by the square of sample variance, 6.25, gives a standardized fourth-power sum of 5.44.
For n=5, the current source’s multiplier is 1.25 and its subtracted correction is eight. Thus 1.25*5.44-8 equals -1.2, matching the independent KURT target.
Example 2: distinguish the unadjusted moment convention
The same five values have population variance 2 and population fourth central moment 6.8. Their unadjusted moment ratio is 6.8/(2^2)=1.7, or excess -1.3 after subtracting three.
Neither 1.7 nor -1.3 equals KURT’s adjusted result -1.2. The difference is a definition and sample-adjustment issue, not an arithmetic contradiction.
Keep the function name and adjustment convention next to a reported statistic. A chart label that says only “Kurtosis” can hide an important difference between calculations.
Advanced application: compare a concentrated dataset with an extreme value
For 1, 1, 1, 1 and 6, the independently expected KURT result is 5. This differs sharply from the evenly spaced dataset despite the same count of five observations.
Fourth powers make the statistic sensitive to large deviations. That sensitivity is why it should not be described merely as a measure of peak height. Tail behavior and individual extreme observations can strongly affect the value.
The function does not identify an observation as a mistake or authorize removing it. Investigate unusual values using the original records rather than changing data simply to obtain a more familiar statistic.
Advanced application: check location and scale invariance
Adding ten to every value in the first dataset gives 11, 12, 13, 14, 15 and the same independent result -1.2. Multiplying all values by two also leaves KURT unchanged.
The standardization cancels a common shift and a nonzero scale factor. Even multiplying by a negative factor leaves fourth-power shape unchanged. These are useful mathematical controls, not a promise that every extreme floating-point magnitude will be equally stable.
Errors and troubleshooting
The current source uses a division-by-zero path for fewer than four usable values or zero standard deviation. A constant range therefore cannot supply this standardized statistic. Replacing the error with zero would incorrectly suggest a normal-reference excess value.
Check the numeric count when selected ranges contain text or blanks. Ignoring a missing value is different from inserting an actual zero, which changes the mean and shape. The documented range behavior does not establish every direct-text or array coercion.
These independent examples do not certify native error messages, extreme-value accuracy or a statistical sampling conclusion. An observed kurtosis value alone does not prove a distribution family, and small-sample estimates can be unstable.
Availability and verification
The exact function is documented in Collabora Office 26.04 desktop help. The cited help snapshot was retrieved on 2026-10-13 and hash-verified October 15, 2026. Desktop documentation does not establish Online or mobile support. First introduction and last functioning release are unknown in this evidence.
No native Collabora Office execution occurred. All numerical results are independent mathematical expectations, with English names and semicolon argument separators. Native error displays, all coercions and extreme numerical behavior are not certified.
Implementation statements are corroborated by Collabora current engine source, retrieved October 15, 2026. This is moving-main source evidence, not a pinned desktop 26.04 binary or execution.