I am currently reading a book about mixture analysis, and in the textbook a posterior for the parameters $\mu1,\mu2,\sigma_1^2,\sigma_2^2$ of a two-component gaussian mixture is derived as follows (S and y are the classifications and data and are given):

$p(\mu1,\mu2,\sigma_1^2,\sigma_2^2|S,y)=p(\mu_1|\sigma_1^2,y,S)p(\sigma_1^2|y,S)\cdot p(\mu_2|\sigma_2^2,y,S)p(\sigma_2^2|y,S)$

where $p(\mu_1|\sigma_1^2,y,S)$ and $p(\mu_2|\sigma_2^2,y,S)$ are normally distributed with some parameters, which I omit for readability here, and $p(\sigma_1^2|y,S)$ and $p(\sigma_2^2|y,S)$ are inverse gamma distributed with some parameters, which I also omit for readability here.

Now my question is: The author states that this is a closed-form solution for the joint posterior $p(\mu1,\mu2,\sigma_1^2,\sigma_2^2|S,y)$. I cannot recognize this product of two subproducts as a known distribution. I also checked the Wikipedia page for conjugate priors, but I could not find anything.

Can someone tell me if I'm overlooking something here?

I already derived the full conditional distributions, so Gibbs sampling would be possible, but this is not a closed-form solution then anymore.

Thanks in advance,



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.