TRIMMEAN in WPS Spreadsheet: Trim Both Tails with Explicit Observation Counts

• Introduced: documented by 2019-05-23 in an official WPS lesson; the first release is not established.
• Availability: the lessons identify WPS Office for Windows Personal, without a numeric build range.
• Scope: exact current supported builds and other WPS clients remain unverified; historical documentation does not establish retirement.
• Checked: October 14, 2026 against function-specific sources. Worked results were independently calculated; no native WPS test was performed.

=TRIMMEAN(array, percent)

Both arguments are required. array is the numeric dataset; percent is the total trimming fraction. The historical Windows lesson documents #NUM! when percent is below 0 or above 1. Its exact percent=1 outcome is not established here, so worked examples stay below 1.

For n numeric observations and a nonnegative fraction p below 1, the documented count rule can be modeled as 2floor(np/2) removed observations. Half are removed from each end. This is an independent explanation of the stated rule, not a separate native formula test.

Examples use English names and comma separators. Keep the source numeric because the lesson does not give a complete mixed-type inclusion table.

Enter these ten numeric scores in A2:A11:

1, 5, 6, 7, 8, 9, 10, 11, 12, 100

=TRIMMEAN(A2:A11,0.2)

Ten times 0.2 is 2, so remove one value from each end: 1 and 100. The retained values are 5, 6, 7, 8, 9, 10, 11 and 12. They total 68, and their expected trimmed mean is 8.5.

For comparison, =AVERAGE(A2:A11) gives 16.9 because it includes all ten values. A lower trimmed result is not automatically more accurate; it answers a different question by deliberately excluding both extremes.

TRIMMEAN does not alter or delete the source cells. The displayed source can remain in its original order; sorting is part of this explanation of which values are removed.

With the same ten values:

=TRIMMEAN(A2:A11,0.3)
=TRIMMEAN(A2:A11,0.4)

At 0.3, the requested count is 3. WPS’s documented rule reduces that to 2, so the expected result remains 8.5.

At 0.4, remove four observations, two from each end: 1, 5, 12 and 100. The remaining six values 6, 7, 8, 9, 10 and 11 total 51 and also average 8.5.

The equal results in this particular dataset do not mean the selected observations are the same. Inspect the retained set as well as the final mean.

On another sheet, enter 0, 1, 2, 3, 4, 5, 6, 7, 8 and 50 in A2:A11.

Evaluate each fraction with =TRIMMEAN(A2:A11,fraction), substituting the numeric fraction. Small percentage changes can leave the excluded count unchanged. This behavior follows the observation count, not a continuous adjustment to each value.

Return to the first dataset. Trimming at 0.2 removes 1 and 100 and averages eight observations.

A separate winsorized dataset would replace 1 with the next value 5 and replace 100 with the next value 12, while keeping ten observations:

5, 5, 6, 7, 8, 9, 10, 11, 12, 12

Its arithmetic mean is also 8.5 here, but the procedures and denominators differ. Do not infer equivalence from this symmetric interior.

For an asymmetric control, use 0, 1, 2, 3, 10 and 100. Removing one observation per end leaves 1, 2, 3 and 10, mean 4. Replacing the extremes instead gives 1, 1, 2, 3, 10 and 10, mean 4.5. TRIMMEAN’s documented rule removes observations; it does not replace them with boundary values.

Use AVERAGE on an explicitly copied retained set to audit the result. Tied boundary values are repeated observations; the count rule does not remove every occurrence merely because one boundary value is excluded.

Native empty-input, error-cell, numeric-text and exact percent=1 results remain unverified. The percentage policy should be chosen before inspecting whether it produces a preferred answer.

• WPS Academy: TRIMMEAN, 2019-05-23 (Chinese; Windows Personal).
• Editorial testing standards.

Expected outputs are independent calculations, not recorded native results. Current-build limits beyond the cited documentation remain unverified.