VARPA Function (OpenOffice Calc)

• Introduced: First release not established; documented by OpenOffice.org 3.3.
• Supported in: Apache OpenOffice 4.1.16; this identifies a verified release, not a final supported version.
• Scope: Desktop Calc, English function names and semicolon-separated formulas.
• Checked: October 14, 2026. Eight reference formulas and the listed diagnostics were recalculated from generated native ODF in 4.1.16 on Linux x86-64, Build 9816. Source and independent reference calculations were also checked.

VARPA calculates population variance after including text as zero and logical values as 1 or 0. It divides the sum of squared deviations by n, the number of included observations. One included observation has zero variance; an input with no included observations is different. Release entry.

=VARPA(Value1 [; Value2; ...])

Supply numeric values, logical values, text, cell references or ranges. TRUE contributes 1 and FALSE contributes 0. Text contributes 0, including the directly supplied numeric-looking string “7”; it is not converted to seven. A genuinely empty referenced cell is ignored. A cell holding text and a cell with no content are therefore different inputs.

Enter these values at the exact addresses before using the range examples. F7 and F8 must be text; enter an apostrophe before 7 to force F8 to text. Enter TRUE() and FALSE() as logical formulas or use logical cell values. Leave the specified cells genuinely empty.

F2:F8 therefore contributes the seven numbers 1, 3, 5, 1, 0, 0, 0. G2:G5 contributes only 2, 4, 6; its final blank is not an extra observation.

The native column records the saved 4.1.16 results. The independent values use high-precision mathematics; displayed decimals are rounded to twelve places. The data setup above is required for range formulas.

If 2, 4 and 6 are the entire population you intend to describe, their mean is 4 and their population variance is 8/3≈2.666666666667. =VARPA(G2:G5) gives that value, ignoring the blank G5. If those readings instead estimate variability in a wider population from a sample, the n−1 estimator answers a different question.

A logical flag population can legitimately use zero-one coding. H2:H4 is 1, 0, 1, with proportion 2/3 and population variance (2/3)×(1/3)=2/9. Preserve logical types and do not replace missing flags with text unless a zero observation is intentional.

The mixed F2:F8 range has seven included values and centered squared sum 152/7. Dividing by seven gives 152/49≈3.102040816327. Excluding the text values would change that calculation; zero and missing are different modeling choices.

For a zero-one population with proportion p, VARPA equals p(1−p), with maximum 1/4 at p=1/2. The native TRUE/FALSE diagnostic returns 0.25. This is a descriptive identity and does not assert that observations are statistically independent. A population variance remains in squared units; use its square root when an output in the measurement unit is required.

=VARPA(1) and =VARPA(F7) return 0, because each supplies one included observation. =VARPA(H5) returns #DIV/0! because H5 is empty and supplies none. The empty argument list also errors. Direct “7” contributes zero, so the pair “7”, 2 has population variance 1. Large offsets and near-equal measurements require more numerical care than these small examples establish.

These additional formulas were recalculated from the same generated ODF input. H5 is a deliberately empty cell where referenced. Error displays are retained as errors, even when an ODS numeric attribute stores zero.

varp, var, stdevp, AVERAGEA.

• Function-specific release entry
• Argument/parser helper
• A-function text-as-zero dispatch

The generated ODF input produced the informational load warning “Not all attributes could be read.” After it was acknowledged, the sheet was hard-recalculated and saved separately. Every formula identity, typed input, input label and tested result was reconciled. True empty cells were distinguished from text and logical values. This does not certify every style attribute. The separately prepared XLSX input was not executed. A deliberately stale SUM cache changed from 999999 to 3, and an uncached SUM returned 7 after hard recalculation. Native evidence covers the eight reference formulas and listed diagnostics only. Other explanatory compositions are independent mathematical checks; other releases, locales and extreme-value behavior were not established by this run.