RANDLANDAU Function (Gnumeric)
Generate a random variate from Gnumeric's standard Landau distribution, with no-argument syntax, heavy-tail cautions, and honest random-output checks.
Verification status: See the evidence and test scope described in this guide. How formula examples are checked
RANDLANDAU generates one random variate from Gnumeric’s standard Landau distribution. It is a sampling function, not the LANDAU density evaluator. A density height and a random observation have different meanings even though the function names share the same distribution.
Syntax and parameter scope
=RANDLANDAU()
There are no arguments: no location, scale, seed, sample-size or cumulative flag. Each call returns one numeric draw. The retained source delegates to Gnumeric’s Landau generator, derived from GNU Scientific Library/CERN routines. Its parameterization must be checked before comparing samples with another package’s convention.
The distribution is asymmetric and has a long positive tail. Do not assume a normal-distribution model, a finite normal-style mean/variance summary, or a small upper bound merely because many observed values look moderate.
Reproducible test with an invariant
A particular numeric output from =RANDLANDAU() is not reproducible by formula alone. The bounded test therefore uses =ISNUMBER(RANDLANDAU()), whose expected result is TRUE for an ordinary generated value. This checks that the runtime recognizes and calculates the function; it does not verify the entire distribution or prescribe the next random draw.
To inspect a small sample yourself, enter =RANDLANDAU() in A2 and fill down a modest range. Label the column “Random draws” and record when it was recalculated. Do not copy a displayed example number from a guide and expect every row or future run to equal it. This guide intentionally provides no fabricated fixed draw as an expected answer.
Independent distribution reference
The GNU Scientific Library reference defines the same standard Landau density through an integral. At x=0, an independent numerical integration gives a density near 0.178854160675; this is a property of the theoretical distribution, not a typical draw or a lower/upper bound. Density checks and random-sample checks should not be conflated.
For a model explicitly defined as location + scaleĂ—X with positive scale and X following this standard distribution, a transformed draw can be constructed by that arithmetic expression. Verify the intended parameterization and units first; RANDLANDAU itself does not interpret those additional model parameters.
Advanced simulation cautions
Heavy-tailed samples can contain unusually large positive values and can make naive sample-mean summaries unstable. Inspect quantiles and the model’s assumptions rather than rejecting a large draw merely because it is inconvenient. A small sample that looks smooth does not establish correctness of a simulation model.
Recalculation can change random results. Save the particular realization as values if it must be reviewed later, with the formula-based model retained separately. There is no worksheet seed argument here, so identical results across sessions or versions are not promised. It is not a cryptographic randomness tool.
Availability and evidence
NEWS records RANDLANDAU’s addition in Gnumeric 1.1.7. The inspected source is retained official 1.12.62, originally obtained October 13–14, 2026. A bounded native 1.12.57 numeric-type invariant was checked on October 17; it is not a goodness-of-fit test. Last functioning release and historical execution are not established. No GUI screenshot or native 1.12.62 execution is claimed.
Native verification
The native 1.12.57 numeric-type check returned TRUE in one recorded execution; no specific draw or distribution-quality certification is implied.
Sources and related guides
- Official Gnumeric function catalog
- Pinned official function help and implementation
- Official release history
- AVERAGE and MEDIAN are sample summaries, not distribution generators.
- GNU Scientific Library Landau definition.