The negative binomial distribution generalises the geometric distribution: it gives the probability that the r-th success happens on trial x.
P(X = x) = C(x − 1, r − 1) pʳ (1 − p)^(x − r), mean = r/p
Example: a salesperson closes 40% of calls. The probability that the 3rd sale comes on the 7th call is 0.1244.
Relationship to other distributions
- With r = 1 it reduces to the geometric distribution.
- It is the waiting-time counterpart of the binomial: the binomial fixes the number of trials, while the negative binomial fixes the number of successes.
- In data analysis, the negative binomial is also used to model overdispersed counts, whose variance exceeds the mean.