Let $X$ be the number of trials preceding the $k$th success in a sequence of independent Bernoulli trials each with probability of success $p$.

Then $X$ has a negative binomial distribution with p.m.f. $p(x) = \binom{x+k-1}{k-1}p^{k}(1-p)^{x}$.

Please explain this pmf using a combinatorial argument. I get why we need to choose $k-1$ successes, but the rest of the pmf is eluding me.


It appears that $x$ is the number of failures before the $k$-th success. Thus, there are $x$ failures and $k-1$ successes before the $k$-th success, for a total of $x+k-1$ trials. There are $\binom{x+k-1}{k-1}$ ways to choose where in the string of $x+k-1$ outcomes the $k-1$ successes fall. The probability of any particular one of these strings is the product of the probabilities of the individual outcomes; $x$ of those outcomes are failures and have probability $1-p$, and $k-1$ are successes and have probability $p$, so the probability of any one of these $\binom{x+k-1}{k-1}$ possible sequences of outcomes is $p^{k-1}(1-p)^x$. The total probability of a string of $x$ failures and $k-1$ successes is therefore


The probability that any of these sequences is followed by a success (namely, the $k$-th success) is $p$, so the probability that the $k$-th success is preceded by $x$ failures is


| cite | improve this answer | |
  • $\begingroup$ Thank you for pointing out the crucial error: I had gotten the definition wrong, defining $X$ as the number of trials (instead of the number of failtures) preceding the $k^{th}$ success. Now, it's easy peasy to get the pmf. $\endgroup$ – Ryan Jul 25 '13 at 20:15
  • $\begingroup$ @Ryan: I thought that that might be the source of the difficulty. You’re welcome. $\endgroup$ – Brian M. Scott Jul 25 '13 at 20:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.