Probability Mass Function, Moments and Factorial Moments of the Negative Binomial Distribution NB$(k,r)$

Abstract

The negative binomial distribution NB$(k,r)$ of Type I is the probability distribution for a sequence of independent Bernoulli trials (with success parameter $p\in(0,1)$) with $r$ nonoverlapping success runs of length $\ge k$. We present a new, more concise, expression for its probability mass function. We show it can also be succinctly written using hypergeometric functions. We also present new expressions (combinatorial sums) for its moments and factorial moments, as opposed to only the mean and variance (which are already known). Next, we present an alternative non-combinatorial viewpoint, which yields expressions for the factorial moments not only for nonoverlapping success runs, but also for runs with an overlap of $\ell$, where $\ell\in[0,k-1]$. The case $\ell=k-1$ is the negative binomial distribution NB$(k,r)$ of Type III. The results also yield the solution for the negative binomial distribution NB$(k,r)$ with a minimum gap between the success runs (explained in the text). Addendum 1/23/2024: The probability mass function and factorial moments are derived from the probability generating function. Addendum 1/26/2024: Alternative expressions are presented for the negative binomial distribution NB$(k,r)$ of Type II.