8
$\begingroup$

Let $\pi$ be the target distribution on $(\mathbb{R}^d,\mathcal{B}(\mathbb{R^d}))$ which is absolutely continuously wrt to the $d$-dimensional Lebesgue measure, i.e :

$\pi$ admits a density $\pi(x_1,...,x_d)$ wrt to $\lambda^d$ with $$\lambda^d(dx_1,...,dx_d) = \lambda(dx_1) \cdot \cdot \cdot \lambda (dx_d)$$

Let us assume that the full conditionals $\pi_i(x_i|x_{-i})$ from $\pi$ are known. So the transition kernel of the Gibbs-Sampler is clearly the product of the full conditionals from $\pi$.

Is the transition kernel absolutely continuously wrt to the $d$-dimensional Lebesgue measure too ?

$\endgroup$
  • 1
    $\begingroup$ i am so confused about the chapter of convergence properties of the gibbs sampler written by casella and Robert. sry for this question it is rather obvious but i need to be sure because it is for my master thesis $\endgroup$ – user2016445 Feb 6 '15 at 16:06
  • $\begingroup$ sorry about our chapter confusing you...! $\endgroup$ – Xi'an Feb 6 '15 at 17:11
  • 2
    $\begingroup$ You're fortunate to have one of the authors answering your question. $\endgroup$ – Glen_b Feb 6 '15 at 17:28
5
$\begingroup$

If you write down the transition of the systematic Gibbs sampler kernel, you get $$\mathbb{P}(X'\in A_1\times\cdots\times A_d|X=x)=\int_{A_1} \pi_1(x'_1|x_{-1})\Big\{\int_{A_2} \pi_2(x'_2|x'_1,x_{-1:2})\cdots \left\{\int_{A_d} \pi_d(x'_1|x_{-d}')\lambda(\text{d}x_d')\right\} \cdots\lambda(\text{d}x_2')\Big\}\lambda(\text{d}x_1')$$ for any product set $A_1\times\cdots\times A_d\in\mathcal{B}(\mathbb{R^d})$ and therefore $$K(x,x')=\pi_1(x'_1|x_{-1})\times\pi_2(x'_2|x'_1,x_{-1:2})\times\cdots\times\pi_d(x'_d|x_{-d}')$$ is the density of a probability measure that is absolutely continuous against the Lebesgue measure on $(\mathbb{R}^d,\mathcal{B}(\mathbb{R^d}))$.

$\endgroup$
  • 2
    $\begingroup$ that is really funny :). thank you again now i feel very comfortable about my chapter for the convergence properties of the gibbs sampler. i really want to thank you for the chapter of convergence properties for the metropolis-hastings ! the minimal necessary conditions are brilliant and i really write down a beautiful proof for the irreducibility of the corresponding markov chain of the MH-Algo. $\endgroup$ – user2016445 Feb 6 '15 at 17:49

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.