Suppose \begin{align} \begin{pmatrix} X_i\\ Y_i \end{pmatrix} \sim_{iid} N \begin{pmatrix} \begin{pmatrix} \mu_1 \\ \mu_2 \end{pmatrix}, \begin{pmatrix} \sigma^2_1 & \rho \sigma_1 \sigma_2 \\ \rho \sigma_1 \sigma_2 & \sigma_2^2 \end{pmatrix} \end{pmatrix} \end{align} with $i=1,...n$, $\sigma_1, \sigma_2 >0$, $\mu_1,\mu_2 \in \mathbb{R}$, $\rho \in (-1,1)$.

I'm trying to calculate an asymptotic distribution of $\rho^{mle}$.

I already know that $\sqrt{n}(\rho^{mle} - \rho) \rightarrow N(0, (1-\rho^2)^2)$ but I failed to derive that variance.

To get the fisher information of $\rho$, I use

$log f(X,Y;\rho) = -\frac{1}{2} log(1-\rho^2) - \frac{1}{2(1-\rho^2)}\Big(\big(\frac{X-\mu_1}{\sigma^1}\big)^2 + \big(\frac{Y-\mu_2}{\sigma^2}\big)^2 \Big) + \frac{\rho}{(1-\rho^2)}\Big(\frac{(X-\mu_1)(Y-\mu_2)}{\sigma_1\sigma_2}\Big)$

to calculate second derivative of that function and then have an expectation.

$\frac{\partial^2 logf}{\partial \rho^2} = \frac{(1- \rho^4)}{(1-\rho^2)^3} - \frac{1+3\rho^2}{(1-\rho^2)^3}\Big(\big(\frac{X-\mu_1}{\sigma^1}\big)^2 + \big(\frac{Y-\mu_2}{\sigma^2}\big)^2 \Big) + \frac{2\rho(1-\rho^2) + 4\rho(1+\rho^2)}{(1-\rho^2)^3}\Big(\frac{(X-\mu_1)(Y-\mu_2)}{\sigma_1\sigma_2}\Big)$

When I use that $E\Big(\big(\frac{X-\mu_1}{\sigma^1}\big)^2 + \big(\frac{Y-\mu_2}{\sigma^2}\big)^2 \Big) = 2$ and $E\Big(\frac{(X-\mu_1)(Y-\mu_2)}{\sigma_1\sigma_2}\Big) = \rho$ to caculate $-E\big[\frac{\partial^2 logf}{\partial \rho^2}\big] = \frac{1+\rho^2}{(1-\rho^2)^3}$.

I can't get what I want. Where did I make a mistake?



Your Answer

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

Browse other questions tagged or ask your own question.