Consider a pair of independent random variables $(V,Z)$ where $Z$ is standard normal. Now suppose that the following equality holds: for a given $a>0$ \begin{align} \mathbb{P}[ a V\le Z| V+Z]=\mathbb{P}[aV \ge Z| V+Z] \text{ a.s.} \end{align}

Question: Can we show that the above is true if and only if $V=\frac{1}{\sqrt{a}}Z'$ where $Z'$ is standard normal independent of $Z$?

To show the forward direction note that if we choose $V=\frac{1}{\sqrt{a}}Z'$, we get \begin{align} \mathbb{P} \left[ \sqrt{a} Z'- Z \le 0| \frac{1}{\sqrt{a}}Z'+Z \right]=\mathbb{P} \left[\sqrt{a} Z'- Z \ge 0| \frac{1}{\sqrt{a}}Z'+Z \right] \text{ a.s.} \end{align}

Now note that $ \sqrt{a} Z'- Z$ and $\frac{1}{\sqrt{a}}Z'+Z$ are independent, to see this observe that \begin{align} \mathbb{E} \left[(\sqrt{a} Z'- Z)(\frac{1}{\sqrt{a}}Z'+Z)\right]=0, \end{align} and recall that uncorolated is equivalent to independence for gaussian random variables.

Consequently, we have that \begin{align} \mathbb{P}[ \sqrt{a} Z'- Z \le 0]=\mathbb{P}[\sqrt{a} Z'- Z \ge 0] \end{align} which is true by summary.

The other direction appears to be hard or not true. I am not sure.
I was thinking of re-writing the equality condition as \begin{align} \mathbb{P} \left[ \left[\begin{array}{c} a\\-1 \end{array} \right] (V,Z) \le 0 | \left[\begin{array}{c} 1\\1 \end{array} \right] (V,Z) \right]=\mathbb{P} \left[ \left[\begin{array}{c} a\\-1 \end{array} \right] (V,Z) \ge 0 | \left[\begin{array}{c} 1\\1 \end{array} \right] (V,Z) \right], \end{align} and then either using some kind projection theorem or change of variable.



Your Answer

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