1
$\begingroup$

I'm a little bit confused about what the meaning of multinomial distributions, at least from what i've gleaned from the wikipedia page on Multinomial Distribution. In essence, a multinomial distribution is the generalized form of a binomial distribution. That is, outcomes are independent, however there are k possible outcomes, each with k success, which gives the probability mass function:

$Mult(n, p_1, p_2, p_3, ..., p_n) = {n \choose x_1, x_2, ... x_n} p_1^{x_1} ...p_n^{x_n} $

However, shouldn't evaluating only $X_1$ give a binomial distribution since the multinomial is just a generalization of it? But throwing that in, we get: $Mult(n, p_1) = {n \choose x_1} p_1^{x_1}$ which can't possibly turn into $Bin(n, p) = {n \choose x} p^x(1-p)^{n-x}$

Am I missing something in my understanding of a multinomial distribution? And if so, how do we get the binomial from the multinomial equation then?

$\endgroup$
3
$\begingroup$

Your formula for the multinomial distribution is a little off: there should be $k$ different $p$s and $x$s with $\sum_{i=1}^k x_i=n$, so that the key coefficient is ${n}\choose{x_1,x_2,\ldots,x_k}$. (You can confirm this with Wikipedia.)

To get the binomial distribution, take $k=2$, $x_1=x$, $p_1=p$ and $p_2=1-p$. Then you have ${n}\choose{x_1,x_2}$$p_1^{x_1}p_2^{x_2}=$$n\choose x_1$$p^{x_1}(1-p)^{n-x_1}$, as desired.

$\endgroup$
  • $\begingroup$ oh, i think i see what i was missing now... since we only have $p_1$ and $p_2$, $p_2 = 1 - p_1$ by definition. Thanks! $\endgroup$ – Jainathan Leung Nov 13 '12 at 2:50

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.