For a linear process $X_t=\mu+\sum_j\varphi_jW_{t-j}$

where $W_t$ is white noise and $\mathbb E(W_t^4)<\infty$ ,

$$ \begin{pmatrix} \hat\rho(1) \\ \hat\rho(2) \\ \vdots \\ \hat\rho(k) \\ \end{pmatrix} \sim AN \begin{pmatrix} \begin{pmatrix} \rho(1) \\ \rho(2) \\ \vdots \\ \rho(k) \\ \end{pmatrix},\frac{1}{n}V \end{pmatrix} $$

Where $$V_{i,j}=\sum_{h=1}^{\infty}(\rho(h+i)+\rho(h-i)-2\rho(i)\rho(h))\times(\rho(h+j)+\rho(h-j)-2\rho(j)\rho(h))$$

And $AN$ denotes Asymptotic Normal.

  • My first question is , why is the variance of $\rho=\frac{1}{n}V$ , that is , why have we multiplied $\frac{1}{n}$ with $V$ ?

  • How does the formula of $V_{i,j}$ come ?

  • $\begingroup$ It's Theorem 7.2.1. from Time Series: Theory and Methods by P. Brockwell and R. Davis, a very thorough proof is given in section 7.3 there. $\endgroup$ – Zhanxiong Jun 29 '15 at 13:11

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.