Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have the following set of equations: $$ y = \alpha x + \epsilon $$ $$ z=\beta y + \mu $$ and $\epsilon$ is N(0,$\sigma_{1}^2$), $\mu$ is N(0,$\sigma_{2}^2$) and $x$ is N(0,1) and they are all independent from each other. I can easily find the distribution of $z|y$, $y$ and $z$. However, I could not find the distribution of $y|z$ and $y|x,z$. Do you have any ideas?

share|improve this question

1 Answer 1

up vote 1 down vote accepted

Introduce the shorthand $g(t,u)=\exp\left(-t/(2u)\right)$.

The density of the distribution of $y$ is proportional to $g(y^2,\alpha^2+\sigma_1^2)$. The density of $(y,\mu)$ is proportional to $g(y^2,\alpha^2+\sigma_1^2)\cdot g(\mu^2,\sigma_2^2)$. Since $\mu=z-\beta y$, the density of $(y,z)$ is proportional to $$ g(y^2,\alpha^2+\sigma_1^2)\cdot g((z-\beta y)^2,\sigma_2^2)=h(z)\cdot g((y-\nu z)^2,\tau^2), $$ for some constants $\nu$ and $\tau^2$ to be determined and some irrelevant function $h$. Hence $y$ conditionally on $z$ is normal with mean $\nu z$ and variance $\tau^2$.

Algebraically, one must solve $$ \frac{y^2}{\alpha^2+\sigma_1^2}+\frac{(z-\beta y)^2}{\sigma_2^2}=\frac{(y-\nu z)^2}{\tau^2}+k(z), $$ for some function $k$ independent on $y$. This yields $$ \frac1{\tau^2}=\frac1{\alpha^2+\sigma_1^2}+\frac{\beta^2}{\sigma_2^2}, \qquad \frac{\nu}{\tau^2}=\frac{\beta}{\sigma_2^2}, $$ which determine uniquely $\nu$ and $\tau^2$.

The distribution of $y$ conditionally on $(x,z)$ can be computed by the same method.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.