1

How to calculate the following multivariate probability mass function:

$P(X_1-X = n, X_2-X = n, ..., X_{N-1}-X = n)$

Where $n$ and $N$ are positive integers, and $X_i$ and $X$ are iid random variables with the following discrete probability distribution:

$P(i)=\frac{C_{i-1}}{2^{2i-1}}$; $C_{i}$ are Catalan numbers.

Looking at: Integrating pdf times cdf squared and walking backwards, from multivariate to integral representation I got:

$ \sum _{i=0} ^{+ \infty} {P(i)P(i+n)^{N-1}}$

However, it appears that the last formula gives me correct results only for N=2. Am I doing something wrong?

Francesco
  • 75
  • 1
    What have you tried so far? If these are independent random variables, you can work out the answer by conditioning on the value of $X$. – Doctor Milt Sep 25 '23 at 13:05
  • They are iids, I have amended the question. I have also added something I tried but didn't work. – Francesco Sep 25 '23 at 15:50

1 Answers1

0

The start of your solution looks correct to me. We have $$ \begin{align} p & = P(X_1-X = n, \ldots , X_{N-1}-X = n ) \\ & = \sum_{i=1}^\infty P(X_1-X = n, \ldots , X_{N-1}-X = n | X = i) P(X = i) \\ & = \sum_{i=1}^\infty P(X_1 = n+i, \ldots , X_{N-1} = n+i) P(X = i) \\ & = \sum_{i=1}^\infty f_X(n+i)^{N-1} f_X(i) \end{align} $$ by independence, where $f_X$ is the PMF. Substituting in $f_X(i)$ gives $$ p = \sum_{i=1}^\infty \left[\frac{C_{n+i-1}}{2^{2(n+i)-1}} \right]^{N-1} \frac{C_{i-1}}{2^{2i-1}} $$

Doctor Milt
  • 3,056