Hi Guys, Im just doing this question. Ive found the autocorrelation at lag 0 and 1, in my usual way - at lag 0 is the variance, at lag 1 using cov(e_n, e_n-1)... From this i obtained autocorrelations 1-Beta^2 and Beta for lags 0 and 1 respectively. Looking at the solutions each of these should have a theta^2 at the end of it. This has confused me, because previously in chapter 13 question 5, the solutions imply that the variance of a MA process is 1 ie var(e_n)+var(-5e_n-1)+var(6_en-2)=1+5^2_6^2=32. I dont understand the difference here, why do we need theta^2 sometimes, but not others. please can someone help? Thanks!
Hi Molly We always need sigma^2, or whatever we are using to represent the variance of the white noise terms. The solution to 13.5 is wrong, there should be a sigma^2 in the autocovariances. It cancels out when calculating the autocorrelations, but it should really still be there for the gamma's. Thanks for pointing this out! We'll get a correction sorted. Hope this helps Andy