Let $X_1,...,X_n$ be (not necessarily independent) identically distributed random variables with $E[X_1^2]<\infty$. Assume that there is a fixed $h\ge1$ so that $Cov(X_i,X_j)=0$ for $|i-j|\ge h$. Proof that with $S_n:=\sum_{i=1}^nX_i$ $$Var(S_n)\le 2nhVar(X_1)$$ Proof.

We have $$Var(S_n)=\sum_{i=1}^nVar(X_i)+2\sum_{\substack{1\le i < j \le n}} Cov(X_i,X_j)$$ $=$(assumption) $$nVar(X_1)+2\sum_{\substack{1\le i < j \le n \\ h > |i-j|}} Cov(X_i,X_j)$$$\le$ (Cauchy-Schwarz-ineq.)$$nVar(X_1)+2Var(X_1)\sum_{\substack{1\le i < j \le n \\ h > |i-j|}}1$$

and that has to be $\le 2nhVar(X_1)$.

Can anyone explain the last inequality to me?


Since $i<j$, $h > |i-j|$ means $h > j-i$ and hence $i<j<i+h$, hence fixing $i$, $j$ can take at most $h-1$ values.

Hence $$\sum_{\substack{1\le i < j \le n \\ h > |i-j|}}1 \leq n(h-1)$$


$$n+2\sum_{\substack{1\le i < j \le n \\ h > |i-j|}}1 \leq n+2n(h-1) \leq 2n + 2n(h-1)=2nh$$


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.