2
$\begingroup$

$\newcommand{\Var}{\operatorname{Var}}$ $\newcommand{\Cov}{\operatorname{Cov}}$ I wonder how come that the variance $\Var(sX)=s^2\Var(X)$, but $\Var\left(\sum_{i=1}^s X_i\right)=\sum_{i=1}^s \Var(X_i)+\sum_{i=1,j=1 ,i\neq j}^{s}2\Cov(X_i,X_j)=\sum_{i=1}^s \Var(X_i)=s\Var(X)$ for $s$ uncorrelated random variables with equal distribution. What is the main reason for that?

Thank you.

$\endgroup$
  • 1
    $\begingroup$ Your second statement is only true if the variables are uncorrelated. In your first statement, the summed variables are maximally correlated: they're all equal. $\endgroup$ – Chris Eagle Oct 25 '12 at 16:18
  • $\begingroup$ Take $s=2$. $$\text{var}(sX) = \text{var}(X+X) = \text{var}(X)+\text{var}(X)+2\text{cov}(X,X) = 4\text{var}(X)=s^2\text{var}(X)$$ $\endgroup$ – Dilip Sarwate Oct 25 '12 at 16:19
  • $\begingroup$ @ChrisEagle: Thanks, I meant to write it. $\endgroup$ – Ben Benli Oct 25 '12 at 16:21
1
$\begingroup$

$\newcommand{\Var}{\operatorname{Var}}$ $\newcommand{\Cov}{\operatorname{Cov}}$ Assuming that the $X_i$'s are a random sample, it is because $X_i$ and $X_j$ are independent for $i\neq j$. On the other hand, $\Var(sX)=\sum_{i=1}^s\sum_{j=1}^s\Cov(X,X)=s^2\Cov(X,X)=s^2\Var(X)$.

In general, for independent $A$ and $B$, we have $\Var(aA+bB)=a^2\Var(A)+b^2\Var(B)$.

You can read a derivation here: http://en.wikipedia.org/wiki/Variance#Properties.

$\endgroup$
  • $\begingroup$ In this case independence suffices but the weaker hypothesis of uncorrelatedness is enough. $\endgroup$ – Michael Hardy Oct 25 '12 at 16:30

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.