0

Let $Z \in \mathbb{R}^2$ be an i.i.d. Gaussian vector with mean $M$ where $P_{Z|M}$ is it's distribution.

Question: I am looking for an non-trivla example of function $g: \mathbb{R}^2 \to \mathbb{R}$ such that \begin{align} E[g(Z)\mid M=\mu]=0, \forall \mu \in C \end{align} where $C=\{\mu: \frac{\mu_1^2}{r_1^2}+\frac{\mu_2^2}{r_2^2}=1 \}$. That is, $C$ is an ellipse.

Some Thoughts: I was able to find an example of $g$ when $r_1=r_2=r$ that is when $C$ is a circle. For example, let \begin{align} g(x)= \|x\|^2 -(2-r^2) \end{align} Then, \begin{align} E[g(Z)|M=\mu]&= E[\|Z\|^2 \mid M=\mu]-(2-r^2)\\ &=\operatorname{Var}[Z \mid M=\mu]+ \|\mu\|^2-(2-r^2)\\ &=2 +r -(2-r)=0 \end{align}

However, I was not able to find an example of such a function if $C$ is an ellipse but not a circle.

Boby
  • 6,381
  • You say "Where $P_{Z,\mid,M}$ is$,\ldots$", but then you never use that notation is the rest of what you write. Why introduce a notation that you never use? – Michael Hardy Jun 19 '21 at 18:28

1 Answers1

0

Hint

If $\ f(x)=\frac{x_1^2}{r_1^2}+\frac{x_2^2}{r_2^2}\ $, what is $\ E\big[f(x)\,\big|\,M=\mu\big]\ $ for $\ \mu\in C\ $?