NEGBINOMDIST Function (Gnumeric)

• Introduced: First release not established; this name is present in the tested Gnumeric 1.12.57 and the cited 1.12.62 source.
• Availability: Supported in Gnumeric 1.12.62 according to its function-specific registration. No retirement endpoint is established.
• Scope: Desktop Gnumeric with the fn-stat module enabled. Examples use English comma-separated formulas.
• Checked: October 14, 2026. Reference and boundary tables were recalculated by Gnumeric 1.12.57 and checked independently. The 1.12.62 contract was source-checked, not executed.

For positive integer t and nonnegative integer f, the mass is COMBIN(f+t−1,f)×p^t×(1−p)^f. Total trials equal f+t. The distinction matters: this is not the probability of t successes among a fixed total of f trials. Official NEGBINOMDIST reference.

=NEGBINOMDIST(f,t,p)

All arguments shown are required. Use numeric cells or numeric inputs. The formulas use decimal points and commas; localized entry conventions require separate checks. Gnumeric 1.12.62 help and evaluator.

Enter each formula in a separate cell. The displayed references are rounded to twelve decimal places and agree with the recorded Gnumeric 1.12.57 recalculation within ordinary floating-point precision. Keep the stored result for subsequent calculations.

Suppose independent trials succeed with probability 0.5, and the process stops at the third success. Put failures 2 in A2, target successes 3 in B2 and probability 0.5 in C2. =NEGBINOMDIST(A2,B2,C2) gives 0.1875.

The event takes five trials in total. Its last trial must be the third success; the earlier four contain two successes and two failures. This explains the coefficient COMBIN(4,2), rather than COMBIN(5,3).

Sum a bounded failure range

The probability of needing no more than two failures before three successes is =NEGBINOMDIST(0,3,0.5)+NEGBINOMDIST(1,3,0.5)+NEGBINOMDIST(2,3,0.5), giving 0.5. This spelling has no cumulative selector, so make the sum explicit.

For an independent integer-case check, the same event is reaching at least three successes within five trials: =1-BINOMDIST(2,5,0.5,TRUE()). Both give 0.5 for this model. For different parameters, keep the total-trial transformation f+t consistent.

Extra domain guard: Gnumeric rejects f+t−1≤0 before calling its negative-binomial helper. Consequently NEGBINOMDIST(0,1,0.5) gives #NUM! even though the usual geometric model assigns probability 0.5 to that event. Do not extend the ordinary formula to this cell behavior. The wrapper floors f and t and checks p in[0,1]; use integer counts and positive target successes rather than relying on nonstandard boundary combinations. The tested NEGBINOMDIST(2,3,0) also returns #NUM!: the probability helper rejects p=0 even though the outer wrapper allows that endpoint. For ordinary use choose 0<p≤1.

The following are observed cell results from the same 1.12.57 command-line recalculation. F5 was genuinely empty. A blank is not necessarily an omitted argument or a valid observation, so validate required data instead of treating every defaulted zero as intentional.

The fixtures do not certify every locale, extreme tail, imported type or newer binary build. The current wrapper agrees with the documented argument contract; shared numerical helpers can change between releases.

• BINOMDIST
• COMBIN
• NEGBINOMDIST official manual entry
• Version-pinned argument contract and calculation
• Negative-binomial probability helper