Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I could find the mean and variance but I found some difficulties in computing the covariance of the following random variables.

Here is the problem:

Suppose $Y_1,Y_2,\ldots$ are i.i.d RVs with $\mathbb{E}Y_i=\mu$ and Var$(Y_i)=\sigma^2\in(0,\infty)$. Set $X_k:=Y_k-3Y_{k+1}+Y_{k+2},k=1,2,\ldots$, and put $S_n:=X_1+\dots+X_n$. Compute Cov$(X_j,X_k)$ for $j\not=k$.

Here is my attempt:
We let $\tilde{Y_j}:=Y_j-\mu$ (so that $\mathbb{E}\tilde{Y_j}=0$), then we have Cov$(X_j,X_k)=\mathbb{E}(\tilde{Y_j}-3\tilde{Y_{j+1}}+\tilde{Y_{j+2}})(\tilde{Y_k}-3\tilde{Y_{k+1}}+\tilde{Y_{k+2}})$.

I know that we have to separate into cases of different values of $|j-k|$, but I am not sure how to proceed. I am a bit confused in manipulating random variables with indexes. Any helps on the detailed workings will be really appreciated.

Many many thanks!

share|improve this question
    
I believe your approach is correct. It's just tedium; you have to consider all the cases where two R.V.s that are being multiplied with one another are (in)dependent. –  par May 27 '13 at 6:24
    
@par. Yes, you are right. Thanks. –  user71346 May 27 '13 at 10:34
    
So what do you need $S_n$ for in the question? –  wolfies May 27 '13 at 11:59
add comment

2 Answers

up vote 1 down vote accepted

We can assume that $j> k\geq 1$. Using that the covariance is bi-linear we obtain the expression: $$ \begin{align*} \mathrm{Cov}(X_j,X_k)=&\mathrm{Cov}(Y_j,Y_k)-3\mathrm{Cov}(Y_j,Y_{k+1})+\mathrm{Cov}(Y_j,Y_{k+2})-3\mathrm{Cov}(Y_{j+1},Y_k)\\ +&9\mathrm{Cov}(Y_{j+1},Y_{k+1})-3\mathrm{Cov}(Y_{j+1},Y_{k+2})+\mathrm{Cov}(Y_{j+2},Y_k)\\ -&3\mathrm{Cov}(Y_{j+2},Y_{k+1})+\mathrm{Cov}(Y_{j+2},Y_{k+2}). \end{align*} $$ Now, use that $\mathrm{Cov}(Y_i,Y_j)=0$ whenever $i\neq j$ (due to independence). This allows you to deal with seperate cases: $j=k+1$, $j=k+2$ and lastly $j>k+2$.

share|improve this answer
    
Thank you for the help. Just spotted a minor mistake, the coefficient 9 should be positive, and there is no case for $j=k$ since $j\not=k$ is given in the question. But thanks very much, it is very helpful :) –  user71346 May 27 '13 at 10:33
    
@user71346: You're welcome, and thanks for pointing out the mistakes. –  Stefan Hansen May 27 '13 at 10:43
add comment

Start by expanding the two parenthesis under the expectation. Alternatively, save a bit of time by using the fact that covariance is multilinear:

$$\mbox{Cov}(aX+bY,cR+dS)=ac\mbox{Cov}(X,R)+ad\mbox{Cov}(X,S)+bc\mbox{Cov}(Y,R)+bd\mbox{Cov}(Y,S)$$

and that two independent random variables have zero covariance.

share|improve this answer
    
Thank you for the help. –  user71346 May 27 '13 at 10:30
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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