How to get the two above conditional distribution?

43 Views Asked by At

Random variables X and N have joint distribution, defined up to a constant of proportionality,

$f(x,n) \propto \frac{e^{-3x}x^n}{n!}$ for $n=0,1,2,...$ and $x>0$. Note that X is continuous and N is discrete.

The question is implement a Gibbs sampler to sample from this distribution.

But I don't understand the solution :

The conditional distribution of $N$ given $X = x$ is proportional to $x^n/n!$, which is a Poisson distribution with parameter $x$.

The conditional distribution of $X$ given $N = n$ is proportional to $e^{−3x}x^n$, which is a gamma distribution with parameters $r = n + 1$ and $\lambda = 3$.

How to get the two above conditional distribution?

The solution also said

$$p(x,n)=\frac{e^{−3x}x^n}{n!} / \int_0^\infty \Sigma_{n=0}^\infty \frac{e^{−x}x^n}{n!} dx= 2 \frac{e^{−3x}x^n}{n!}$$

I don't understand above. Where does 2 come from?

I can only recognize that $f(x,n)$ has the kernel of Poisson ($3x$).

1

There are 1 best solutions below

0
On BEST ANSWER

Note $$p(x,n)=f(x,n)/\int_{\mathbb{R}}\sum_{n\in \mathbb{Z}_{>0}}f(x,n)dx$$ just normalizes $f(x,n)$ so that it represents a valid joint mass-density and integrates to one.

The conditional mass function of $N$ given $X=x$ is proportional to

$${f(x,n) \over \sum_{n\in \mathbb{Z}_{>0}}f(x,n)}\propto f(x,n)\propto x^n/n!,$$

where $\propto$ in this context only accounts for terms depending on $n$ (taken $X=x$ as fixed).

The conditional density of $X$ given $N=n$ is proportional to

$${f(x,n) \over \int_{\mathbb{R}}f(x,n)dx}\propto f(x,n)\propto e^{-3x}x^n,$$

where $\propto$ in this context only accounts for terms depending on $x$ (taking $N=n$ as fixed).