FREQUENCY Function (Gnumeric)

• Introduction: Gnumeric 0.32. NEWS explicitly lists FREQUENCY among the implemented functions.
• Source support: Registered in the Gnumeric 1.12.62 plugin. No retirement is established. The last functioning application version is unverified for this guide.
• Evidence dates: Official Gnumeric source was retained October 13–14, 2026 and reviewed October 15. The GOffice 0.10.61 source archive was retrieved October 15 and its relevant range routine reviewed the same day. Typed filtering, exact counting and rational interpolation checks are independent reference calculations. No native Gnumeric, compiled-library, headless application or GUI execution was performed.

=FREQUENCY(data_array,bins_array)

Both arguments are required. The first contains observations and the second contains upper cutoffs. After sorting the numeric bins, the first bucket contains values at or below the smallest cutoff. Each later ordinary bucket contains values above the previous cutoff and at or below the current one. The final bucket contains values above the largest cutoff.

Text, Booleans and physically empty cells are ignored in both data and bins. Numeric zero is retained. Input errors propagate during collection. If no numeric bins remain, the callback returns the number of numeric data points as a scalar instead of the usual multi-row array.

This is a count function, not a density or an automatic bin-selection procedure. The cutoffs determine the question. A change in cutoffs can change the frequency table without changing any observation.

Create FrequencyCounts. Put observations 6,10,10,15,20,21,30,30,50 in A2:A10. Leave A11 physically empty, put text 12 in text-formatted A12, and Boolean TRUE in A13. Put numeric upper cutoffs 10,20,30 in C2:C4.

Select E2:E5 and enter =FREQUENCY(A2:A13,C2:C4) using the pinned manual’s array procedure: select the complete output range, enter the formula and press Ctrl+Shift+Enter. This is documented entry guidance, not an observed GUI test or a promise of dynamic spill.

The sum is 9, matching the nine numeric observations. The text 12 does not enter the second bucket and TRUE is not counted as one. Including the overflow row is essential; selecting only three output cells would omit part of the distribution being described.

Array notation in the table describes the input and output values; it is not a localized formula-literal recipe. Enter values into ordinary cells. Source inspection determines ignored types and errors; the integer bucket counts are independently enumerated.

In a copy, change the final numeric observation from 50 to 20. The counts become 3,3,3,0, still summing to nine. The new 20 is included in the bucket ending at 20 because the upper boundary is inclusive.

Restore the original observations. In another copy change the middle cutoff from 20 to 25. The counts become 3,3,2,1: the observation 21 moves into the wider middle interval. This change is caused by the bin definition rather than a change in the data.

If a measurement is exactly on a cutoff, do not count it in both adjacent buckets. The source loop assigns each numeric observation to the first sorted cutoff that is at least as large as that observation, or to the final overflow bucket.

Supplying bins 30,10,20 still gives 3,2,3,1, because Gnumeric sorts the cutoffs before counting. It does not reorder labels you wrote elsewhere on the sheet. Keep labels in ascending cutoff order so the displayed meaning matches the returned array.

With bins 10,20,20,30, the result is 3,2,0,3,1. The repeated cutoff creates an empty-width bucket: observations equal to 20 were already assigned to the earlier cutoff. Removing a duplicate bin changes the output shape even when it does not change the meaningful counts.

Using the original E2:E5 results, divide each count by their total nine. The exact relative frequencies are 1/3,2/9,1/3,1/9, summing to one. These are observed proportions for this dataset, not probabilities guaranteed for a larger population.

The running count totals are 3,5,8,9. Dividing those totals by nine gives cumulative proportions for the successive upper cutoffs and finally all observations. Keep the unbounded overflow row clearly labeled rather than pretending it has an ordinary finite upper endpoint.

Unequal bucket widths require care when using a chart as a density estimate. FREQUENCY returns counts; it does not divide by interval width, choose a chart type or infer the underlying continuous distribution.

If there are no numeric data and three numeric cutoffs, the reference array is four zeroes. If the bins range has only ignored entries, the result is the scalar numeric-data count. If an input contains an error value, collection propagates it rather than treating it as an extra category.

Check that cutoffs and observations are actual numbers, that the output has one more row than the number of numeric cutoffs, and that labels follow sorted order. Native array padding behavior, locale-dependent input conversion and file transfers were not tested. Preserve the raw observations so changes to bins remain auditable.

• Function help and worksheet callback
• Numeric collection flags and iteration
• Gnumeric array-entry manual source
• Gnumeric official function catalogue

COUNT, SUM, MODE-MULT.