Could you help me with the following:

I have that $$T(x):=\frac{X(nx)-E[X(nx)]}{\sqrt{n}} \xrightarrow{d} N(0, \frac{x^k}{k})$$ for each fixed $x>0$, where we also have that $\frac{X(nx)}{t}$ is the empirical Distribution function of a Distribution, where $\frac{E[X(nx)]}{t}$ is the Distribution function.

Then, it is meantioned that this proves that $T(x) \xrightarrow{d} W(\frac{x^k}{k})$, where W is a Standard brownian Motion.

Why? Is there any reference for that? As reference, it is mentioned the following Theorem, which does not help me: Suppose the $\xi_n$ are Independent and have a common Distribution function $F(t)$. Then, with $$Y_n(t,\omega)=\sqrt{n} (F_n(t,\omega)-F(t)),$$ $Y_n \xrightarrow{d} Y$, where $Y$ is the Gaussian random element of $D$ specified by $E[Y(t)]=0$, $E[Y(s)Y(t)]=F(s)(1-F(t))$ for $s\leq t$.

Thus, what really is my question: What has a Standard brownian Motion to do with a Gaussian random element? Could you explain this to me? I would really appreciate any help!!

(DEFINITION STANDARD BROWNIAN MOTION: $W(0)=0$ and $t \mapsto W(t)$ is continuous with probability 1 and $W(t)$ has stationary, jointly Independent increments and the increment $W(t_1+t_2)-W(t_1)$ is normal-distributed $N(0,t_2)$.)


For every nonnegative $u$, the distribution of $W(u)$ is $N(0,u)$ hence, for any parameter $u(x)$, the assertion that $T(x)\stackrel{d}{\to}N(0,u(x))$ and the assertion that $T(x)\stackrel{d}{\to}W(u(x))$ are logically equivalent.

  • $\begingroup$ thank you for your answer. I think, this is exactly what I am looking for. As the reference I am looking at is much more difficult, I wonder, whether you know a good paper where this is explained relatively easy? This would really help me. As I also have a Problem with the Notation (in my paper), is it true that $W(u) =^d N(0,u)$? I am not sure whether this sign $=^d$ means this. $\endgroup$ – user146358 Aug 14 '14 at 13:02
  • 1
    $\begingroup$ In this case it means that the RHS is the distribution of the LHS. $\endgroup$ – Did Aug 14 '14 at 13:05
  • $\begingroup$ Okay, then I think, this should hold, doesn't it? As we have $W(u+s)-W(s) =^d N(0,u)$, we have with $W(0)=0$ that $W(u)=^d N(0,u)$? Or does this not make sense in context of $W$? I am not sure what $W$ exactly is. Does it have a Distribution, so does it make sense to write $W=^d ...$? $\endgroup$ – user146358 Aug 14 '14 at 13:07
  • 1
    $\begingroup$ Wow, you seem really lost here... Yes, $W(u)\stackrel{d}{=}N(0,u)$ by definition. $\endgroup$ – Did Aug 14 '14 at 13:09
  • $\begingroup$ Yes. I didn't ever have a course in probability theory, I never heard of these things. So yes, I am really lost.. :-) Thank you for your help. If you could send me an easy paper, where this is explained, I would really appreciate this. (All I found is with Gaussian random elements, and as you can think, I have no idea what this is). $\endgroup$ – user146358 Aug 14 '14 at 13:11

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.