I'm trying to prove the following property:

$$ var \Big( \frac{1}{n}\sum^n_{i=1} Y_i \Big) = \frac{\sigma^2}{n} + \frac{n-1}{n}\rho\sigma^2 $$

So far, I've gotten up to this point:

$$ \textrm{var}\Big( \frac{1}{n}\sum^n_{i=1}Y_i \Big) = \frac{1}{n^2}\textrm{var}\Big( \sum^n_{i=1}Y_i \Big) \\ = \frac{1}{n^2}\Bigg( \sum^n_{i=1}\textrm{var}(Y_i) + 2\sum^{n-1}_{i=1}\sum^n_{j={i+1}}\textrm{cov}(Y_i, Y_j)\Bigg) \\ $$

where $\rho$ is the correlation coefficient between $Y_i$ and $Y_j$, since

$$ \textrm{var}\Big( \sum^n_{i=1}Y_i \Big) = \sum^n_{i=1}\sum^n_{j=1}\textrm{cov}(Y_i, Y_j) \\ = \sum^n_{i=1}\textrm{var}(Y_i) + \sum_{i \ne j}\textrm{cov}(Y_i, Y_j) \\ = \sum^n_{i=1}\textrm{var}(Y_i) + 2\sum^{n-1}_{i=1}\sum^n_{j={i+1}}\textrm{cov}(Y_i, Y_j) $$

I know that$\sum^n_{i=1}var(Y_i) = n\sigma^2$, but I cannot see how the second term becomes $\frac{n-1}{n}\rho\sigma^2$. Any hints would be greatly appreciated. Thank you!


You must be assuming $var(Y_i) = \sigma^2$ for all i, and $corr(Y_i,Y_j) = \rho$ for all $i \ne j$.

The expansion of $var(1/n*\Sigma_{i=1}^nY_i)$ has $n^2$ terms, of which $n$ are the variance (i = j) terms, and $n^2 - n$ are the $i \ne j$ terms. So your 2nd term becomes $1/n^2*(n^2 -n)*cov(Y_i,Y_j)$

| cite | improve this answer | |
  • $\begingroup$ Hi, Mark. Yes, that's another assumption made. This makes perfect sense. Thank you! $\endgroup$ – user122514 Apr 21 '18 at 21:01

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.