Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Suppose that $Y=A+\epsilon$ where $\epsilon$ is a RV and given some other random variable $\eta$ we have that:

$\epsilon|\eta$ ~ $N(\rho\eta,\sigma^2)$

Suppose I was asked to find $Pr(Y=y|\eta)$ This (I think) is easily equal to: $\frac{1}{\sigma}\phi(\frac{y-A-\rho\eta}{\sigma})$ where $\phi$ is the standard normal pdf.

Now I suppose I need to find $Pr(Y=y|\eta>x)$ The answer I have for this, using conditional probability formula (P(A|B)=$\frac{P(A\cap B)}{P(B)})$is:

$\frac{\int_x^{\infty}\frac{1}{\sigma}\phi(\frac{y-A-\rho z}{\sigma})dz}{P( \eta>x)}=\frac{\Phi(\frac{y-A-\rho x}{\sigma})}{(P( \eta>x))\rho}$

which does not seem right to me.

Edit: I forgot about the y's

share|cite|improve this question
$Y$ is a continuous random variable. Isn't it true that $Pr(Y=y|\eta)=0$? – Patrick Li Jan 30 '13 at 21:31
It wasn't the best choice of notation. I meant I wanted the PDF. – Greg Jan 30 '13 at 22:09
up vote 1 down vote accepted

Obviously, the answer should depend on the distribution of $\eta$. Assume that $\eta$ has PDF $g$ and CDF $G$, then the conditional PDF $f_x$ of $Y$ conditionally on $\eta\gt x$ is such that $$ f_x(y)=\frac1{1-G(x)}\int_x^{+\infty}\frac1\sigma\phi\left(\frac{y-A-\rho t}{\sigma}\right)\,g(t)\mathrm dt, $$ where $\phi$ is the standard gaussian PDF and $$ 1-G(x)=\mathbb P(\eta\gt x)=\int_x^{+\infty}g(t)\mathrm dt. $$

share|cite|improve this answer

Since $Y= A+ \epsilon$, then we know that $\epsilon = Y-A$. Also $P(Y = y) = P(A+\epsilon = y) \Longleftrightarrow P(\epsilon = y-A)$. So it seems that $$P(Y = y| \eta) = P(\epsilon = y-A | \eta)$$ $$= \Phi \left(\frac{y-A-\rho \eta}{\sigma} \right) $$

So $$P(Y = y| \eta > x) = P(\epsilon = y-A| \eta>x)$$

$$= \int_{x}^{\infty} \Phi \left(\frac{y-A-\rho \eta}{\sigma} \right) \ d \eta$$

share|cite|improve this answer
Thanks for pointing that out, I've made the corrections. However, the main question remains. – Greg Jan 30 '13 at 20:23
@Greg: Since we are given the distribution $\epsilon| \eta \sim N(\rho \eta, \sigma^2)$, I don't think we need the extra $\frac{1}{\sigma}$. – proton Jan 30 '13 at 20:42
Careful: $P(Y=y)=0$ and $P(Y=x\mid \eta\gt x)=0$ for every $y$. – Did Jan 30 '13 at 22:03

Your Answer


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.