Let $X_1,...,X_n$ be random variables with identical distribution, and for all $i=1,...,n$ $\mathrm{Var}(X_i)$ exist.
1. Show that the covariance between each two random variables exist.
2. Show that :
$\mathrm{Var} \sum _{i=1}^{n} X_i = n \mathrm{Var} (X_1) + n(n-1)\mathrm{Cov}(X_1,X_2)$

Well, at the first question I tried to split it to two cases: case 1, that $\mathrm{Cov}(X_i,X_j)$ when $i=j$, and in this case it is simply $\mathrm{Var}(X_i)$ which exists.
but then I tried to show the other case, when $i\neq j$, and I dont really know how. Is it correct to use Cauchy-Schwarz inequality $|\mathrm{Cov}(X_i,X_j|^2 \leq \mathrm{Var}(X_i)\mathrm{Var}(X_j)$?

Other then that, I don't know how to show the second question, can someone give me a hint?


Cauchy Schwarz is fine. For the second equation $\def\Var{\mathop{\rm Var}}\def\Cov{\mathop{\rm Cov}}$(which will hold only if not only the $X_i$, but the pairs $(X_i,X_j)$, $i \ne j$ are identically distributed) recall that \begin{align*} \Var\sum_i X_i &= \Cov\left(\sum_i X_i, \sum_j X_j\right)\\ &= \sum_{i,j} \Cov(X_i, X_j)\\ &= \sum_i \Var(X_i) + \sum_{i\ne j} \Cov(X_i, X_j) \end{align*} and the fact the the $X_i$ are identical distributed. This gives $\Var(X_i) = \Var(X_1)$ for all $i$. If we have that the pairs are identically distributed, we have $\Cov(X_i, X_j) =\Cov(X_1, X_2)$ for all $i \ne j$, hence $$ \Var\sum_i X_i = \sum_i \Var(X_1) + \sum_{i\ne j} \Cov(X_1, X_2) = n \Var(X_1) + n(n-1)\Cov(X_1, X_2) $$

  • $\begingroup$ Do I need to show that the pairs $(X_i,X_j) i\neq j$ are identically distributed? How do I do that? It does not follow from the fact that $X_1,...,X_n$ are identically distributed? $\endgroup$ – Buzi Jun 26 '14 at 7:31
  • 1
    $\begingroup$ @Buzi You cannot show this from your assumptions, you have to assume it additionally if you want (2) to hold. It will not follow from $X_1, \ldots, X_n$ being identically distributed: Let $X_1, X_2$ be independent and $N(0,1)$ distributed (to be concrete). Let $X_3 := X_1$. Then $X_1, X_2, X_3$ are identically distributed, but: ${\rm Cov}(X_1, X_2) = 0$ and ${\rm Cov}(X_1, X_2) = {\rm Var}(X_1) = 1$. $\endgroup$ – martini Jun 26 '14 at 7:44
  • $\begingroup$ Oh, you are right, I even got this very own counter example in class. Thanks for your help! You helped me a lot! $\endgroup$ – Buzi Jun 26 '14 at 7:46

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.