2
$\begingroup$

Let $\{\epsilon_t\}$ and $\{\eta_s\}$ two non-correlated stationary process $\forall t,s$. Show that $\{\epsilon_t+\eta_s\}$ is a stationary process with autocovariance function equal the sum of the autocovariances of $\{\epsilon_t\}$ and $\{\eta_s\}$.

Proof:

If $\{\epsilon_t\}$ and $\{\eta_s\}$ are stationary then they have mean and variance constants $\forall t,s$ and the covariance is independent of time. Let

$$\begin{cases}E[\epsilon_t]=\mu_1\\Var(\epsilon_t)=\sigma_1^2\\Cov(\epsilon_t,\epsilon_{t+h})=\rho_1\end{cases}\qquad\text{and} \qquad\begin{cases}E[\eta_s]=\mu_2\\Var(\eta_s)=\sigma_2^2\\Cov(\eta_s,\eta_{s+h})=\rho_2\end{cases}$$

where $\mu_1,\mu_2,\sigma_1,\sigma_2,\rho_1,\rho_2\in \mathbb{R}$

For $\{\epsilon_t+\eta_s\}$

$$E[\epsilon_t+\eta_s]=E[\epsilon_t]+E[\eta_s]=\mu_1+\mu_2$$

$$Var(\epsilon_t+\eta_s)=Var(\epsilon_t)+Var(\eta_s)+2cov(\epsilon_t,eta_s)=\sigma_1^2+\sigma_2^2$$

since that the process are non-correlated then $cov(\epsilon_t,\eta_s)=0$

$$Cov(\epsilon_t+\eta_s,\epsilon_{t+h}+\eta_{s+h})=Cov(\epsilon_t,\epsilon_{t+h})+cov(\epsilon_t,\eta_{s+h})+Cov(\eta_s,\epsilon_{t+h})+cov(\eta_s,\eta_{s+h})$$ $$=cov(\epsilon_t,\epsilon_{t+h})+cov(\eta_s,\eta_{s+h})=\rho_1+\rho_2$$

Then $\{\epsilon_t+\eta_s\}$ is a stationary process.

Since the autovariance function $\gamma_\epsilon(h)=Cov(\epsilon_t,\epsilon_{t+h})$ and $\gamma_\eta(h)=Cov(eta_s,eta_{s+h})$, the autovariace function of $\{\epsilon_t+\eta_s\}$ is $$\gamma_{\epsilon+\eta}(h)=\gamma_\epsilon(h)+\gamma_\eta(h)$$

Is this right?

$\endgroup$
1
$\begingroup$

Your proof looks correct. Two remarks:

  • Even if it is obvious, I think it is god to say that $\epsilon_t+\eta_t$ has a finite second moment.
  • You should mention again the fact that the processes $\epsilon$ and $\eta$ are uncorrelated when you compute $\operatorname{ Cov} (\epsilon_t+\eta_s,\epsilon_{t+h}+\eta_{s+h})$.
$\endgroup$

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.