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.

Let $Y,X$ be jointly normally distributed and assume that they are highly correlated. I'm interested in knowing what happens to the survival function as the variables become perfectly correlated. Specifically, I'm interested in this instance:

$$ \lim_{\rho_{YX} \rightarrow1} \Pr[Y>y,X>\mu_X] $$

I tried getting around the integrals but did not get anywhere. I plotted numerically different examples and found that it does seem to converge.

Below is the plot of $\Pr[Y>y,X>\mu_X]$ (blue) and $\Pr[Y>y]$ (dashed red) for a bivariate normal distribution with $\mu_Y=\mu_X=0$, $\sigma_Y=\sigma_X=1$ and $\rho_{YX}=0.999$. As it can be seen, the former converges to the latter for $y>0$. How can I prove this and get the general result?

Example of survival function at limit

share|improve this question
add comment

1 Answer

up vote 2 down vote accepted

When $\mu_X=\mu_Y=0$ and $\sigma_X=\sigma_Y=1$, the random vector $(X,Y)$ can be realized as $(X,Y)=(\varrho Y+\sqrt{1-\varrho^2}Z,Y)$ where $Z$ is standard normal and independent of $Y$. Then, $A_\varrho=[Y\gt y,X\gt0]$ is $A_\varrho=[Y\gt y,Z\gt-a(\varrho)Y]$ with $a(\varrho)=\varrho/\sqrt{1-\varrho^2}$. Since $a(\varrho)\to+\infty$ when $\varrho\to+1$, $[Z\gt-a(\varrho)y]\to\Omega$ if $y\gt0$ and $[Z\gt-a(\varrho)y]\to\varnothing$ if $y\lt0$. Thus, $P[A_\varrho]\to P[B]$ where $B=[Y\gt y,Y\gt0]=[Y\gt\max(y,0)]$. This explains the blue curve.

In the general case, $(X,Y)=(\mu_X+\sigma_X\varrho T+\sigma_X\sqrt{1-\varrho^2}Z,\mu_Y+\sigma_YT)$ where $Z$ and $T$ are standard normal and independent. Then, $A_\varrho=[Y\gt y,X\gt\mu_X]$ is $$ A_\varrho=[T\gt(y-\mu_Y)/\sigma_Y,Z\gt-a(\varrho)T]. $$ Since $a(\varrho)\to+\infty$ when $\varrho\to+1$, $[Z\gt-a(\varrho)t]\to\Omega$ if $t\gt0$ and $[Z\gt-a(\varrho)t]\to\varnothing$ if $t\lt0$. Thus, $P[A_\varrho]\to P[B]$ where $B=[T\gt(y-\mu_Y)/\sigma_Y,T\gt0]$, that is, $B=[T\gt\max((y-\mu_Y)/\sigma_Y,0)]$. This yields a translation by $\mu_Y$ of a dilation by $\sigma_Y$ of the blue curve.

share|improve this answer
add comment

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.