5
$\begingroup$

Let $\{X_1,X_2,\ldots\}$ be random variables with mean $\mu$ and variance $\sigma^2$ both finite.

I know that the CLT gives

$$\frac{\sqrt{n}}{\sigma}\left(\frac{1}{n}\sum_{i=1}^nX_i - \mu\right)\overset{d}{\rightarrow}N(0,1)$$

But is it true that

$$\frac{\sqrt{n}}{\sigma}\left(\frac{1}{n}\sum_{i=1}^nX_i\right)\overset{d}{\rightarrow}N(\mu,1)$$

It seams right, but it cant prove it.

$\endgroup$
3
  • 2
    $\begingroup$ 1. What makes you say "it seems right"? Can you explain the intuition? 2. Split the LHS of the first line into two terms (the first being the LHS of your second line) and look closely at the second term ... $\endgroup$
    – Glen_b
    Commented Nov 18, 2016 at 6:19
  • $\begingroup$ Got it, it's actually not true! Thx for your comment. $\endgroup$
    – Mur1lo
    Commented Nov 18, 2016 at 6:27
  • $\begingroup$ If you'd care to put a less informal explanation as an answer, that might be good. $\endgroup$
    – Glen_b
    Commented Nov 18, 2016 at 6:29

1 Answer 1

3
$\begingroup$

A simple rearrangement of terms proposed by @Glen_b shows that my intuition was wrong. Also a simple way to see why this is not true is to suppose that $X_i\sim N(\mu,\sigma^2)$. Then all we have to do is note that

$$\frac{\sqrt{n}}{\sigma}\left(\frac{1}{n}\sum_{i=1}^n{X_i}\right) \sim N\left(\frac{\sqrt{n}}{\sigma}\mu,1\right)$$

and that would never converge to $N(0,1)$.

$\endgroup$
1
  • 2
    $\begingroup$ It can't really be said to converge to anything at all, because the distribution of $\sqrt n \bar{X}$ shifts further and further to the right as $n$ increases, beyond every finite bound. $\endgroup$
    – Glen_b
    Commented Nov 18, 2016 at 10:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.