Find The Posterior of $(\mu, \frac{1}{\sigma^2})$

56 Views Asked by At

I am given the following sample:

$6.56, 6.39, 3.30, 3.03, 5.31, 5.62, 5.10, 2.45, 8.24, 3.71,\\ 4.14, 2.80, 7.43, 6.82, 4.75, 4.09, 7.95, 5.84, 8.44, 9.36$

which is a sample from a $N(μ, σ^2)$ distribution.

I am also give that:

  • The prior $μ|σ^2 \stackrel{}{\sim} N(3, 4σ^2)$
  • $σ^2 \stackrel{}{\sim} Gamma(1, 1)$

I am being asked to determine the posterior of $\large(\mu, \frac{1}{\sigma^2})$. How do I find that? Any help is much appreciated!

1

There are 1 best solutions below

0
On BEST ANSWER

Let's set $\theta_1=\mu$ and $\theta_2=\frac{1}{\sigma^2}$ so that $\theta_2$ indicates the precision of a normal distribution.

Your model is the following

$$p(x|\theta_1,\theta_2)=\sqrt{\frac{\theta_2}{2\pi}}e^{-\frac{\theta_2}{2}(x-\theta_1)^2}$$

I think that the Gamma(1;1) is the prior distribution of $\theta_2$ thus the prior

$$h(\theta_1,\theta_2)=\sqrt{\frac{1}{2\pi}}\theta_2^{\frac{1}{2}}\cdot e^{-\frac{\theta_2}{2}(\theta_1-3.4)^2}\cdot e^{-\theta_2}$$

is a Normal Gamma distribution with the following parameters_

  • $\alpha=1$

  • $\beta=1$

  • $\lambda=1$

  • $\mu=3.4$

Now, as usual, multiplying the prior and the likelihood, after some algebraic manipulations, you will find that the posterior distribution is still a Normal Gamma with the following parameters:

  • $\mu_{post}=\frac{3.4+n\overline{X}_n}{n-1}$

  • $\lambda_{post}=n+1$

  • $\alpha_{post}=1+\frac{n}{2}$

  • $\beta_{post}=1+\frac{1}{2}\Sigma_i(X_i-\overline{X}_n)^2+\frac{n(\overline{X}_n-3.4)^2}{2(n+1)}$


Concluding, after seeing the sample you posted, the posterior is

$$ \bbox[5px,border:2px solid red] { h(\theta_1,\theta_2|\mathbf{x})=\sqrt{\frac{21}{2\pi}}\frac{43.14^{11}}{10!}\cdot \theta_2^{10.5}\cdot e^{-10.5\theta_2(\theta_1-8.70)^2}\cdot e^{-43.14\theta_2} \ } $$