I am very confused about the following:

Let us define $VaR_u(Y)$ as $\inf \{l:F_Y(l)\geq\ u \}$ where $F_Y$ is the CDF of the random variable Y.

Suppose $A $ and $X$ are independent random variables and suppose we have $VaR_u(AX)=5$. Furthermore, suppose we are interested in the following probability $P(AX>VaR_u(AX)|A=a)$. It seems as if there are two ways to deal with this probability, but surely there must be only one.

First we can take the stance that $VaR_u(AX)$ is a fixed quantity, namely 5. Thus we have $$P(aX>5)=P(aX>VaR_u(AX))$$

On the other hand one could argue that $VaR_u(AX)$ is dependent on the value of $A=a$ such that we have $$P(aX>VaR_u(aX))=1-u$$

I am inclined to believe that the first interpretation is the right one, but how does one 'choose'?


1 Answer 1


The first interpretation is correct.

$VaR_u(Y)$ is the quantile function. It is a function of $u$ and of the true underlying distribution of $Y$. It does not depend on any particular value of $Y$ or any samples drawn from the true underlying distribution.

For example, if $u = \frac{1}{2}$, $VaR_u(Y)$ is the median of the underlying distribution of $Y$. If you draw samples from some distribution, it does not change the median of that distribution.

To avoid this confusion, you might consider using a different notation, such as $VaR(u, F) = \inf\{l:F(l)\geq u\}$.

  • $\begingroup$ Thank you so much for clarifying the dependency $VaR_u(Y)$ has on $u$ and the distribution function of Y. This has been bothering me for two days. $\endgroup$
    – Joogs
    May 21, 2017 at 18:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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