Why when I compute the autocovariance function of a non-zero mean AR(1), X(t)-u=Φ(X(t-1)-u)+ε the presence of the mean does not change my result and so the formula should be the same of a zero-mean AR(1)? I've tried to do that, but i obtain a extra-term (the yellow one), so there is something wrong!

enter image description here

  • $\begingroup$ Could you show where the "$Variance(X_{t-1})$ term comes from? I believe that in introducing it you have double-counted a term equal to $-\phi u^2.$ $\endgroup$ – whuber Dec 19 '18 at 17:28
  • $\begingroup$ @whuber The variance of X(t-1) is the result of the expectation of X(t-1) squared that I can write as variance $\endgroup$ – White Noise Dec 19 '18 at 17:34
  • $\begingroup$ Haven't you assumed $E[X_{t-1}]=u$?? Unless $u=0,$ the expectation of $X_{t-1}^2$ is not its variance. $\endgroup$ – whuber Dec 19 '18 at 17:37
  • $\begingroup$ @whuber You re perfectly right. And so, what does X(t-1) squared stand for? $\endgroup$ – White Noise Dec 19 '18 at 17:41

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.