Is it possible to determine the numerical value of the parameters of a moving average process just by looking at the correlogram (Autocorrelation function and Partial Autocorrelation function) of the model?

Suppose we have a simple $MA(1)$ process: \begin{equation} y_{t} = \epsilon_{t}+\alpha\epsilon_{t-1} \end{equation} with $\epsilon_{t} \sim WN(0, \sigma^{2}_{\epsilon})$.

Just by looking at the Autocorrelation function and Partial Autocorrelation function of the process without estimating a proper regression, can we derive the value of $\alpha$?


1 Answer 1


I computed the correlation between $y_{t}$ and $y_{t-1}$ as follows: \begin{equation} Cov(y_{t}, y_{t-1})=E[y_{t}y_{t-1}]-E[y_{t}]E[y_{t-1}]=E[y_{t}y_{t-1}] = E[(\epsilon_{t}+\alpha \epsilon_{t-1})(\epsilon_{t-1}+\alpha \epsilon_{t-2})] = E[\alpha \epsilon_{t-1}^{2}] = \alpha \sigma^{2}_{\epsilon} \end{equation} \begin{equation} Var(y_{t})=Var(y_{t-1})=\sigma^{2}_{\epsilon}+\alpha^{2}\sigma^{2}_{\epsilon}=\sigma^{2}_{\epsilon}(1-\alpha^2) \end{equation} Therefore: \begin{equation} Corr(y_{t}, y_{t-1})=\frac{Cov(y_{t}, y_{t-1})}{Var(y_{t})} = \frac{\alpha}{1+\alpha^2} \end{equation} From wich I can get the value of alpha.

Thank you for the intuition!


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.