3
$\begingroup$

Starting from $Var(\overline{x})$ I am trying to algebraically show that it is equal to $\frac{\sigma^2}{N}$ using the fact that the variance of the sum equals to the sum of variances. I start by $$\operatorname{Var}(\overline{x})=\operatorname{Var}\left(\frac1N\sum_{i=1}^Nx_i\right)$$ then $$\frac1N\sum_{i=1}^N \left(\frac1N \sum_{i=1}^Nx_i-\mu\right)_i^2 = \frac1N \sum_{i=1}^N \left[\frac1{N^2} \left(\sum_{i=1}^N x_i\right)^2-\frac{2\mu}{N}\sum_{i=1}^N x_i+\mu^2 \right]_i$$ which becomes $$\frac1N\sum_{i=1}^N \left[\overline{x}^2-2\mu \overline{x}+\mu^2\right]_i = \frac1N \sum_{i=1}^N(\overline{x}-\mu)_i^2=\frac{\sigma^2}{N}$$

$\endgroup$
  • 1
    $\begingroup$ what is $M$? What is the first expression on 2nd line? $\endgroup$ – bdeonovic Aug 13 '16 at 13:15
  • $\begingroup$ what is $\bar{x}_N$? $\endgroup$ – David C Aug 13 '16 at 13:29
  • $\begingroup$ Sample mean where a sample has $N$ values of the random variable. $\overline{x_N}=\frac1N\sum_{i=1}^Nx_i$ $\endgroup$ – ahra Aug 13 '16 at 13:34
  • 1
    $\begingroup$ How is $\bar{x}$ different than $\bar{x}_N$? $\endgroup$ – David C Aug 13 '16 at 13:42
  • $\begingroup$ How do you go from the right hand side of the 1st line to the left-hand side of the 2nd line? And how do you go from the middle of the 3rd line to the right-hand side of the 3rd line? Those don't make any sense at all, thereby rendering your proof incorrect. You made one error with the first of these, and you made an offsetting error with the second of these. $\endgroup$ – Mark L. Stone Aug 13 '16 at 14:37
2
$\begingroup$

$$ \begin{align*} Var(\bar{X}) &= Var\left(\dfrac{1}{n}\sum_{i=1}^n X_i\right)\\ &= \dfrac{1}{n^2}Var\left(\sum_{i=1}^n X_i\right) \end{align*}$$ variance of sum is equal to sum of variances because $X_i$ are independent $$ \begin{align*} &= \dfrac{1}{n^2}\sum_{i=1}^n Var(X_i) \\ &= \dfrac{n\sigma^2}{n^2}\\ &= \dfrac{\sigma^2}{n} \end{align*} $$

$\endgroup$
  • $\begingroup$ I think the proof I posted is perhaps more fundamental because it doesn't use any properties of variance like Bienaymé's formula. $\endgroup$ – ahra Aug 13 '16 at 12:57
  • $\begingroup$ @ahra But you have. Otherwise where are all the $\sum_i\sum_j$ terms in the second step of your derivation? $\endgroup$ – tchakravarty Aug 13 '16 at 13:12
  • $\begingroup$ Well I did use the fact that $X_i$ since I'm studying this in context of experimental errors, but I don't see where I used any variance properties. Regarding the sum multiplication terms I'm not sure exactly where should they appear? $\endgroup$ – ahra Aug 13 '16 at 13:26
2
$\begingroup$

It doesn't really seem like your proof makes sense, but it might just be that you are skipping steps, which is making it difficult to understand. Here is a more complete proof of $Var(\bar{X}) = \sigma^2/n$

$$ \begin{align*} Var(\bar{X}) &= E\Big( (\bar{X} - \mu)^2 \Big) \\ &= E\Big( \big(\dfrac{1}{n}\sum_{i=1}^n X_i - \mu\big)^2 \Big) \\ &= E\Big( \big(\dfrac{1}{n}(X_1+\ ...\ + X_n) - \mu\big)^2 \Big) \\ &= E\Big( \big(\dfrac{1}{n}\big((X_1-\mu)+\ ...\ + (X_n-\mu)\big)\big)^2 \Big) \\ &= \dfrac{1}{n^2}E\Big( \big((X_1-\mu)+\ ...\ + (X_n-\mu)\big)^2 \Big) \\ &= \dfrac{1}{n^2}E\Big( \sum_{i=1}^n \sum_{j=1}^n(X_i - \mu)(X_j-\mu) \Big) \\ &= \dfrac{1}{n^2}\sum_{i=1}^n \sum_{j=1}^n E\Big((X_i - \mu)(X_j-\mu)\Big) \\ \end{align*}$$ Note, however, that since the $X_i$s are independent, then if $i\neq j$, then $E((X_i - \mu)(X_j - \mu)) =Cov(X_i,X_j) = 0$. Thus: $$\begin{align*} &= \dfrac{1}{n^2}\sum_{i=1}^n E\Big((X_i - \mu)^2\Big) \\ &= \dfrac{1}{n^2}\sum_{i=1}^n Var(X_i) \\ &= \dfrac{1}{n^2}\sum_{i=1}^n \sigma^2 \\ &= \dfrac{1}{n^2} n\sigma^2 \\ &= \dfrac{\sigma^2}{n} \end{align*}$$

$\endgroup$

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.