1
$\begingroup$

I am trying to evaluate $\mathbb{E}[\operatorname{sign}\langle v,z\rangle]$ for $v\in\mathbb{R}^n$ fixed and $z_i\sim N(0,1)\ \forall\ i\in[n]$.

The $\operatorname{sign}$ part is what is confusing me. Clearly, $$ \mathbb{E}[\operatorname{sign}\langle v,z\rangle] = \mathbb{E}\Big[\operatorname{sign}\sum_{i=1}^nv_iz_i\Big] = \mathbb{P}\Big[\sum_{i=1}^nv_iz_i > 0\Big] - \mathbb{P}\Big[\sum_{i=1}^nv_iz_i < 0\Big]. $$

But I don't know how to simplify further (i.e. to get the expectation just on the $z_i$). I am wondering how to pass the $\operatorname{sign}$ operator through the expectation.

$\endgroup$
1
$\begingroup$

Observe that $z$ and $-z$ have the same distribution.

Consequently $\operatorname{sign}\langle v,z\rangle$ has the same distribution as $\operatorname{sign}\langle v,-z\rangle=-\operatorname{sign}\langle v,z\rangle$, so that:

$$\mathbb{E}[\operatorname{sign}\langle v,z\rangle]=\mathbb{E}[\operatorname{sign}\langle v,-z\rangle]=\mathbb{E}[-\operatorname{sign}\langle v,z\rangle]=-\mathbb{E}[\operatorname{sign}\langle v,z\rangle]$$

This implies that: $$\mathbb{E}[\operatorname{sign}\langle v,z\rangle]=0$$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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