TRIMMEAN in Calligra Sheets: Trimmed Average, Examples and Limits

• Introduced: Documented by 26.08.2; original introduction not established.
• Availability: Supported in Calligra Sheets 26.08.2, based on the function-specific registration.
• Scope: The official application listing provides Linux installation; other packaged builds are unverified here.
• Checked: October 14, 2026. Documentation/source check; no new native test in Calligra Sheets 26.08.2.

TRIMMEAN calculates an average after trimming a matching number of low and high observations from a numeric dataset. Its second argument is the combined fraction of observations targeted for removal across both tails. KDE’s XML describes this outlier-reduction purpose and gives the syntax TRIMMEAN(dataSet; cutOffFraction). KDE handbook · pinned XML.

=TRIMMEAN(A2:A9;0.25)

If A2:A9 contains the exact eight values 12, 7, 20, 7, 15, 30, 12 and 4, the independently calculated expected result is 12.1666666667. Sorting gives 4, 7, 7, 12, 12, 15, 20, 30; with a total cutoff fraction of 0.25, one smallest value and one largest value are removed. The six remaining values sum to 73, so 73 ÷ 6 = 12.1666666667. This is mathematical/source-reasoning verification, not a result from a native Calligra run.

Important implementation detail: The inspected Calligra 26.08.2 C++ code calculates its per-tail cut as floor(dataSet.count() × fraction / 2) and then sorts converted numeric values. As a consequence, a cutoff fraction can be nonzero while zero items are actually removed; the cutoff is in whole observations, equally on both sides. Pinned function registration · pinned implementation.

=TRIMMEAN(dataSet; cutOffFraction)

Both parameters are required. There is no optional default or third argument selecting lower-only/upper-only trimming. The C++ registration calls setParamCount(2) and setAcceptArray(). KDE’s English formula instructions use semicolon argument separators, an initial equals sign, and decimal periods in displayed examples. Other locales may represent the decimal fraction or separator differently; locale behavior has not been tested.

Do not enter 25 when you mean 25%. The argument value for a 25-percent cutoff is 0.25 (or a properly numeric 25%-formatted cell), not 25. The source rejects fractions outside [0,1), but its exact application-visible error text is unknown without native testing.

1. Open a blank Calligra Sheets document and enter Observation into A1, +10 into B1, and ×2 into C1.
2. Enter these eight numeric values in A2:A9 as displayed below. In B2 type =A2+10; in C2 type =2*A2; fill each formula down through row 9.
3. In D1 type Cutoff fraction, in D2 type the numeric value 0.25, and in E1 type Trimmed mean.
4. In E2 enter =TRIMMEAN($A$2:$A$9;D2). With $A$2:$A$9, the range stays fixed if the formula is copied; see KDE’s relative/absolute reference guide.
5. Compare to the independent expected result 12.1666666667. A displayed result may be rounded by cell formatting; increasing shown decimals is not the same as changing the underlying arithmetic.

The original untrimmed sum is 107 and its ordinary average is 13.375. The lower and upper extremes are 4 and 30. Removing those two values leaves 7, 7, 12, 12, 15, 20, with a total of 73 and a trimmed mean of about 12.16666667. Trimmed mean implementation.

All expectations below were computed independently for the exact numeric worksheet. They are not application runtime measurements. The number in the third column is the count removed from each tail for the given fraction, assuming the eight numeric source values.

For the 8-value examples, floor(n × fraction / 2) gives the whole number of items to remove from each side. Duplicates count as separate observations, and the fraction is not rounded up to force removal. KDE’s actual C++ algorithm sorts its numeric list and uses this cutoff for the loop; these are source observations and independent calculations, not a native test report.

Consider a separate fully numeric range G2:G9 containing 6, 7, 7, 8, 8, 9, 10, 100, one value per cell. The usual average is 19.375, because the sum is 155. With a fraction of 0.25, TRIMMEAN removes 6 and 100, and the six retained values sum to 49; their average is 8.1666666667.

AVERAGE provides the untrimmed comparison. Trimming is appropriate only where discarding extremes is part of your analytical method. In small samples it may discard legitimate measurements, and it does not diagnose the reasons an outlier occurred. KDE XML function purpose.

For repeatable analysis, keep the original numbers, record the fraction in a dedicated cell, label the rule for how many observations were removed, and compare trimmed and untrimmed means. Do not quietly replace raw data with trimmed results.

Count of values vs shape of a range. The pinned implementation obtains its cut number from dataSet.count() before independently converting the input range into a sorted numeric list. The underlying Value::count implementation counts stored nonempty array elements, whereas the statistical array helper iterates array cells and converts entries. Consequently, the exact trimming behavior with blank, string, logical or error-valued cells is not established by the all-numeric examples. The code does not contain a visible “ignore all nonnumeric cells” filter here. Native tests are required.

Invalid fractions. Values below zero or greater than/equal to one reach an explicit internal errorVALUE() branch in the function implementation. This is a code-level control-flow observation; it does not prove an exact UI error label or how localized messages are displayed. Avoid trying to extrapolate Excel or Google Sheets error outputs to Calligra.

Empty data. The implementation returns an internal value error if the derived numeric-list count is zero, but there can also be range conversion/aggregation behavior outside the function body. The guide makes no promise about the user-visible error for a completely blank selected range.

Rounding and overflow. The trimmed mean should be compared within the precision shown by the spreadsheet. Source calculations use floating-point accumulators, so retaining excessive digits or combining extremely large values may reveal floating-point precision limitations; no stress tests were performed.

KDE’s formula-entry help describes entering functions, using semicolons, accessing the Function dialog and anchoring references.

The exact review target is Calligra Sheets’ 26.08.2 source revision 72123cc643f6625351139e7df294544255a1056f, associated with KDE Gear’s October 8, 2026 release. The function is documented in the English KF6 handbook and registered in the pinned statistical module, so the intended syntax and algorithm can be investigated. Official announcement · pinned source registration · XML. KDE’s Calligra download page describes Linux distribution packages/Flatpak and ongoing Windows work; this review cannot certify formula availability across platforms, editions or downstream builds.

Compatibility is evidence-based, not assumed: the mathematical idea of a trimmed mean exists in other spreadsheets and statistical packages, but another product’s handling of text, blank cells, percentage rounding, error values, or optional parameters is not evidence for Calligra. No Excel/Sheets/LibreOffice compatibility test, import/export test or native application run was completed. For native validation, test numeric-only baseline; threshold fractions; errors; mixed input types; an empty range; locale separators; differing installed release packages; and visible UI error messages.

• Calligra handbook: TRIMMEAN and pinned XML — purpose, documented syntax and relationship to averages.
• Calligra 26.08.2 function registration, numeric-array helper and TRIMMEAN implementation — argument count, array handling, cut formula and bounds.
• Calligra Value::count() — supports the explanation of array-count behavior.
• KDE formula entry and references, Gear 26.08.2 announcement, and Calligra downloads — syntax and version/platform limits.
• AVERAGE — the untrimmed alternative.

Native limitations: Application evaluation and localization checks have not been performed. No screenshots or native tests are claimed.