LOGINV Function (LibreOffice Calc)

• Introduced: First release not established. The LibreOffice 3.3 function appendix documents this name; historical presence does not prove introduction.
• Supported in: LibreOffice 26.8.1.1. No final supported release has been established.
• Scope: Desktop Calc. Current argument defaults are stated for the cited release, not inferred for every older version.
• Checked: October 14, 2026. Official documentation/source and independent mathematics checked; LibreOffice Calc was not run for this guide.

LOGINV finds the positive cutoff x for which the cumulative LOGNORMDIST result is p. Its formula is exp(μ+σΦ⁻¹(p)). Mean μ and StDev σ describe LN(X), not the arithmetic mean and SD of the positive variable X. Official reference.

The median is exp(μ), so changing σ while keeping μ fixed does not change the median. Other percentiles move: increasing σ spreads the distribution and raises an upper percentile while lowering a lower one. The arithmetic mean, exp(μ+σ²/2), is generally greater than the median.

=LOGINV(Probability [; Mean [; StDev]])

LibreOffice 26.8.1.1 accepts one through three arguments. Omitted Mean defaults to 0 and omitted StDev to 1. Therefore =LOGINV(0.5) also returns 1. The release evaluator establishes that argument contract.

Use 0<Probability<1 and StDev>0. A percentage cell holding 90% supplies 0.9; the number 90 does not. Both probability endpoints are rejected as invalid arguments in the pinned source. Mean may be negative: it is a log-scale location, so a negative μ can still give a positive output.

Enter each formula in a separate numeric cell. The expected mathematical values are independently calculated and rounded to twelve decimal places.

Put 0.9 in A2, log-mean 0 in B2 and log-SD 1 in C2. In D2 enter =LOGINV(A2;B2;C2). The expected 90th percentile is approximately 3.602224479279. A2 is a probability, while D2 is a measurement; use different number formats and labels for them.

In E2 enter =LOGNORMDIST(D2;B2;C2;1). It should recover approximately 0.9. The same parameters must be used on both sides. Do not round D2 before this check or substitute the density mode of LOGNORMDIST.

For an arithmetic mean 10 and arithmetic SD 5, log-SD is SQRT(LN(1+(5/10)^2)), approximately 0.472380727077, and log-mean is LN(10)−log-SD²/2, approximately 2.191013317337. Put those calculations, rather than their rounded displays, in B2:C2 in the correct order.

The median is then approximately 8.944271909999, while the 90th percentile is approximately 16.385447242959. Calculate each with LOGINV using the full log parameters. The fact that the median differs from the arithmetic mean is expected for this skewed model.

A lognormal model cannot describe negative values or a point mass at zero without a separate modeling choice. The function does not determine whether observed data follow that distribution. It only calculates the quantile of the parameters supplied.

Near p=1, exp(μ+σΦ⁻¹(p)) can become extremely large. Preserving the original probability matters: a displayed 100% may still be an interior value, but replacing it with the literal 1 changes the request to an invalid endpoint. Extreme outputs, input coercion and locale-specific entry require installed-application checks beyond these ordinary examples.

When comparing with another product, check its argument count first. A matching function name does not establish matching defaults or mean/SD interpretation.

• LOGINV documentation
• LibreOffice 26.8.1.1 evaluator and argument contract