Let a autoregressive process $AR(1)$ given by $$x_t=\sum_{j=0}^\infty \phi^jw_{t-j}$$ where $|\phi|<1$ and $w_{t-j}$ is i.i.d a white noise with mean 0 and variance $\sigma^2$.

The autocovariance function is $$\gamma(h)=cov(x_{t+h},x_t)=E\Big[\Big(\sum_{j=0}^\infty \phi^jw_{t+h-j}\Big)\Big(\sum_{k=0}^\infty \phi^k w_{t-k}\Big)\Big]$$ $$=E[(w_{t+h}+\dots+\phi^hw_t+\phi^{h+1}w_{t-1}+\dots)(w_t+\phi w_{t-1}+\phi^2 w_{t-2}+\dots)\qquad\textbf{(1)}$$ $$=\sigma^2\sum_{j=0}^\infty \phi^{h+j}\phi^j=\sigma^2\phi^h\sum_{j=0}^\infty \phi^{2j}=\frac{\sigma_2\phi^h}{1-\phi^2},\qquad h\geq 0$$

I know that last equality is the power series representation of that sum, but I'm having hard time to understand how they get $$\sigma^2\sum_{j=0}^\infty \phi^{h+j}\phi^j$$

from $\textbf{(1)}$. I know that they are using $$E[w_{j}w_{k}]=E[w_t^2]=Var(w_t)=\sigma^2$$ in the expectations, but I don't understood how to get the exponents in $\phi$

Anyone can help me?


Because $w_i$ is a white noise, $cov[w_i,w_j]=0$ when $i\ne j$.

It follows that $E[w_iw_j]=0$ when $i\ne j$.

When you expand formula (1) you can get rid of all the terms $w_iw_j$ where $i\ne j$.

The terms you are left with all have an offset of $h$ in the exponent of $\phi$, i.e. $(1)=E[(\phi^hw_t)(w_t)+(\phi^{h+2}w_{t-1})(\phi w_{t-1})+...]$


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.