1
$\begingroup$

Compute the cross-correlation between the time series

$$X_t = X_{t-1}+w_t$$ $$Y_t = \theta X_t+u_t$$

where $w_t$ and $u_t$ are white noise processes independent of each other.

I have problem when calculating the covariance (the numerator of cross-correlation); here's my step: $$E[(x_t -\mu _{x_t})(y_t-\mu_{y_t})]$$ $$=E[(x_{t-1}+w_t-0)(\theta x_t+u_t+0)]$$ $$=E[x_{t-1}\theta x_t+x_{t-1}u_t+w_t \theta x_t+w_tu_t]$$ $$=0+0+0+0$$ $$=0$$ ??

$\endgroup$
3
  • $\begingroup$ Why did the means $\mu_{x_t}$ and $\mu_{y_t}$ disappear in the first equality and mysteriously reappear in the second equality? Why did you expand $x_t$ in terms of $x_{t-1}$ in that equality? $\endgroup$
    – whuber
    Feb 3 '20 at 19:11
  • $\begingroup$ @whuber it should be u, typo (edited). And i just expand it like a quadratic formula. $\endgroup$
    – GarlicSTAT
    Feb 3 '20 at 19:17
  • $\begingroup$ But that's not what you have done. The way the means slip in and out of your equations makes it impossible to follow what you're doing and makes one (reasonably) suspect your work. Thus, your algebra is the first thing you should check: you need to get that right before you worry about anything else. $\endgroup$
    – whuber
    Feb 3 '20 at 20:28
1
$\begingroup$

Hint: For some strange reason, even though you are computing the cross-correlation at time $t$, you have decided to use the recursive equation to replace $X_t$ with $X_{t-1}$. Instead of doing that, try looking at time $t$, by taking $\mathbb{Cov}(X_t,Y_t) = \mathbb{E}(X_t \cdot (\theta X_t + u_t)).$

$\endgroup$

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.