Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Why is the negative binomial distribution defined as $$P(X=x|r,p)= \binom{x-1}{r-1}p^{r}(1-p)^{x-r}$$

Basically this is the probability that $x$ Bernoulli trials are needed for $r$ successes. So we need $r-1$ successes in the first $x-1$ trials. Then success on the $r^{th}$ trial happens with probability $p$. Why can't we write it as the following:

$$P(X = x|r,p) = \binom{x}{r}p^{r} (1-p)^{x-r}$$

This means that you have $r$ successes in the first $x$ trials.

share|improve this question

2 Answers 2

You wrote that $x$ trials are needed for $r$ successes. That means in particular that $x-1$ trials were not enough. So we hit our goal of $r$ successes at the $x$-th trial.

Suppose for example that we are tossing a fair coin until we get the first head. What is the probability that $2$ tosses are needed for $1$ success? Here $x=2$ and $r=1$. Two tosses are needed precisely if we get TH. This has probability $1/4$. By way of contrast, the probability of exactly one head in two tosses is $1/2$.

share|improve this answer
    
@David Mitra: Thanks for spotting the missing not! Fixed. –  André Nicolas Jan 27 '12 at 20:34

One reason the negative binomial distribution is written that way is that $$\sum_{x=r}^\infty \binom{x-1}{r-1}p^{r}(1-p)^{x-r} =1$$ while $\sum_{x=r}^\infty \binom{x}{r}p^{r}(1-p)^{x-r} \gt 1$ and so is not a probability distribution.

The cause of this is non-mutually exclusive events: if you take a sequence of trials, then that sequence can include both 4 trials and 2 successes, and (if the fifth trial is then a failure) 5 trials and 2 successes; you only get mutually exclusive events if you stop when you first have 2 successes.

An alternative approach would be to look at $Y$ the number of failures before the $r$th success with $\Pr(Y=y) = \binom{y+r-1}{r-1}p^{r}(1-p)^{y}$ noting $$\sum_{y=0}^\infty \binom{y+r-1}{r-1}p^{r}(1-p)^{y} =1.$$

You can of course look at the binomial distribution, taking the sum over $r$ (with $x$ fixed) using your formulation and getting $$\sum_{r=0}^x \binom{x}{r}p^{r}(1-p)^{x-r} = 1.$$

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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