0
$\begingroup$

I have a confusion between two statements of the CLT.

1) Let $X_1,...$ be iid random variables with mean $\mu$ and finite variance $\sigma^2$ then $$\frac{X_1+X_2+...+X_n-\mu \color{red}{n}}{\sigma \sqrt n} \implies N(0,1)$$ where $\implies$ means convergence in distribution.

2) If $X_1,...$ are iid random variables with mean $\mu$ and finite variance $\sigma^2$ then $\bar{X}$ is approximately $N(\mu,\sigma^2)$ for large $n$. (Usually $n>30$).

From 2) We have for large $n$ $$\frac{X_1+X_2+...+X_n}{n} $$ is distributed $N(\mu, \sigma^2)$ then normalising gives

$$\frac{X_1+X_2+...+X_n-\mu}{n\times(\sigma/\sqrt n)}=\frac{X_1+X_2+...+X_n-\mu \color{red}{(1)}}{\sigma \sqrt n} $$ is approx $N(0,1)$.

There is a $n$ missing from the expression of 1) in what I get for 2) (indicated with a red $n$ and then a red $1$ where the $n$ should be).

What is going on here.

$\endgroup$
  • $\begingroup$ In $2)$ $\bar{X_n}$ actually have $\mathcal{N}(\mu, \frac{\sigma^2}{n})$ distribution $\endgroup$ – Mesmerized student May 1 '17 at 12:06
  • $\begingroup$ @Mesmerizedstudent No, the distribution of $\bar X$ is not normal in general. $\endgroup$ – Did May 1 '17 at 12:31
  • $\begingroup$ @Did yes, you are right of course. $\endgroup$ – Mesmerized student May 1 '17 at 12:49
2
$\begingroup$

I'll write $Y_n$ for $(X_1+\cdots+X_n)/n$. Its mean is $\mu$ and its variance is $\sigma^2/n$. Its normalisation is $$\frac{Y_n-\mu}{\sigma/\sqrt n}=\frac{nY_n-n\mu}{\sqrt n\,\sigma}=\frac{X_1+\cdots+X_n-n\mu}{\sqrt n\,\sigma}.$$

Now, what was the problem again?

$\endgroup$
0
$\begingroup$

The first expression in your question is evidently equivalent with:$$\frac{\overline{X}-\mu}{\sigma/\sqrt n}\implies N(0,1)$$ You will arrive there by dividing numerator and denominator of LHS by $n$.

On LHS we recognize the standardizing of $\overline{X}$

Btw, note that $\text{Var}\overline{X}=\sigma^2/n$ so that it is incorrect to state that the distribution of $\overline{X}$ is approximately $N(\mu,\sigma^2)$.

$\endgroup$

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.