VARA in Collabora Office: Sample Variance with Text-Encoded Zeros

• Introduced: original Collabora release and launch date are not established by the reviewed sources.
• Availability: currently documented in Collabora Office 26.04 desktop help.
• Platform scope: online and mobile deployments are not inferred from this desktop-help release.
• Checked: October 14, 2026. Function-specific documentation and independent calculations; no native Collabora execution.

=VARA(argument1;[argument2];...)

Collabora documents up to 255 arguments and at least two supplied values. The arguments can be values or references. Its explicit text rule supports the ordinary nonnumeric labels used here; it does not establish every Boolean, numeric-text or error case.

The numeric audit model divides the encoded squared-deviation total by n minus one. Variance has squared units. Taking a square root would produce the different standard-deviation measure.

Enter 3, 9 and Pending in A2:A4:

=VARA(A2:A4)

Expected result: 21. Encoding Pending as zero gives 3,9,0, with mean 4. Squared deviations are 1,25,16, totalling 42. Dividing by two gives 21.

The explicit scalar formula =VARA(3;9;"Pending") has the same independent expectation under the documented text rule. This is an encoding example, not a claim that the unavailable measurement truly equals zero.

For just numeric 3 and 9, the mean is 6 and squared deviations total 18. The sample variance is 18, using denominator one.

Thus including the text-encoded zero changes the result from 18 to 21. The related STDEVA model would take a square root; it would not return 21 in the original units.

Add text Review in A5. Encoded values become 3,9,0,0, with mean 3. Squared deviations 0,36,9,9 total 54.

=VARA(A2:A5)

Expected result: 18, from 54/3. The squared-deviation total increased from 42, yet the sample variance decreased because the denominator also changed.

This counterexample explains why a missing-data policy should be documented directly rather than inferred from whether the final statistic rose or fell.

Restore the three-cell range and replace Pending with numeric 6. The values 3,9,6 have mean 6 and squared-deviation total 18, giving expected sample variance 9.

Keep a count of actual measurements separately from encoded statuses. A later report with variance 9 represents three numeric observations, while the earlier result 21 included one text-zero placeholder.

For the original encoded 3,9,0 data, population variance is 14, from 42/3. Sample VARA gives 21, from 42/2. Their ratio is 3/2 for this three-value fixture.

=VARA(3;9;0)

Expected result: 21. This literal-zero control makes the text encoding visible without changing the arithmetic. VARPA describes the population interpretation.

A genuinely empty cell and a supplied text label are different input states. The concise VARA section does not establish that they share a result. Formula-empty strings, numeric text, logical values and errors remain outside these controls.

Use the smallest numeric/text fixture when diagnosing a changed imported result, then inspect actual stored types. Do not replace all calculation failures with zero: zero variance has a meaningful interpretation for included equal values.

• Collabora Office 26.04: VARA.
• Editorial testing standards.

Complete function-specific documentation reviewed. Worked outputs are independent mathematical or explicitly documented type-rule controls, not application screenshots or native test results. The stated unresolved boundaries retain their scope.