I am trying to calculate the auto-covariance of time series $Z_t$. Given a weakly stationary process $Y_t$; $t \in \mathbb{Z}$: $$Z_t=a+Y_t$$ Now, my goal is to show that $Z_t$ is also a stationary process. I know that $\mathbb{E}[Y_t]$ should be constant, and it is, $\mathbb{E}[Y_t] = a + m$. Then, I need to prove that the auto-covariance $\newcommand{\cov}{\rm Cov} \cov(Z_t, Z_{t+h})$ depends only on $h$.

I am trying to prove it by calculating the covariance as: $$\cov(Z_t, Z_{t+h}) =\mathbb{E}[Z_tZ_{t+h}] - (m+a)^2$$ but I am not getting anywhere. I am not sure what should I show.

I hope I was clear. If you have any insight please let me know.

  • $\begingroup$ Do you have any distributional assumption for $Y_t$? You mentioned the expectation, but what about variance of $Y_t$ ? $\endgroup$ – Sheikh May 3 '15 at 18:22
  • $\begingroup$ I know about the stationary status of $Y_t$, hence I know that its expectation and variance are constant, they do not depend on $t$. $\endgroup$ – user11291 May 4 '15 at 18:00

I assume your goal is to show that $Z_t$ is also a weakly stationary process.

First, you showed ${Z_t}$ has a constant mean:

$$ \mu(Z) = E(Z_t) = E(a + Y_t) = a + E(Y_t) = a + \mu(Y), $$

where $\mu(Y)$ is constant given that $Y_t$ is weakly stationary.

Then, you show that $\gamma_k(Z)$ the covariance the process of ${Z_t}$ only depends on the lag $h$, and not on time $t$:

$$ \gamma_h(Z) = cov(Z_t, Z_{t+h}) = cov(a + Y_t, a + Y_{t+h}) = cov(a, a) + cov(a, Y_{t+h}) + cov(Y_t, a) + cov(Y_t, Y_{t+h}) = cov(Y_t, Y_{t+h}) = \gamma_h(Y), $$

where the last equality is true given that $Y_t$ is weakly stationary.

I believe the answer is complete, unless somebody can point out something missed out.


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.