1
$\begingroup$

Here is the problem (not homework),

Let $U_1,\cdots,U_n$ be i.i.d. uniform$(-n,n)$ random variables. For $-n<a<b<n$, we set $1_{U_i}(a,b)$ be the indicator function such that $1_{U_i}=1$ if $U_i\in(a,b)$ and 0 otherwise. What is approximate distribution as n large of $U_1+,\cdots,+U_n$.

I computed the characteristic function of $U_1+,\cdots,+U_n$, i.e., $\phi(t) = \left(\frac{sin(nt)}{nt}\right)^n$, but I don't know how to get the final result. By the way, I have no idea to use the hint in the problem. Please provide me some hints or references. Thanks!

$\endgroup$
  • $\begingroup$ First, do the I read this correctly that you have $n$ instances of this random variable $U$, and also that $-n$ and $n$ are the bounds of the uniform distribution that the $U$s are instances of? $\endgroup$ – Peter Ellis Sep 11 '12 at 10:14
  • 2
    $\begingroup$ Second, I don't see how the a, b, and the indicator function for $U$ being between them are relevant to the distribution of the sum of the $U$s. What is the reason for thinking they are relevant? $\endgroup$ – Peter Ellis Sep 11 '12 at 10:16
  • $\begingroup$ Hint 1: Central limit theorem. Hint 2 (R): n=1000; sim = numeric(10000); for(i in 1:10000) sim[i]=sum(runif(n,-n,n)); hist(sim); hist(rnorm(10000,0,n^1.5/sqrt(3)),add=T) $\endgroup$ – user10525 Sep 11 '12 at 10:27
  • $\begingroup$ @Procrastinator PeterEllis, thanks for your help. Since $Var(U_1 +\cdots+U_n) = 2n^3 \to \infty$ as $n \to \infty$, I am not pretty sure I can use the central limit theorem directly. $\endgroup$ – Jim Sep 11 '12 at 11:52
  • 1
    $\begingroup$ @Jim There is no need for calculating the variance of the sum. You just need to know the variance of a single variable. The approximation is hidden in my second hint. This is a $\mbox{Normal}(\mu=0,\sigma=n^{3/2}/\sqrt{3})$. Try to check the CLT and verify the details. $\endgroup$ – user10525 Sep 11 '12 at 11:58
2
$\begingroup$

Putting aside the questions in my comments above (which might mean I have misunderstood completely), my hint is -

in general, what is the approximate distribution of the sum of $n$ independent and identically distributed random variables for large $n$?

$\endgroup$

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.