Looking further at sum of Gaussian random variables will look like a Brownian motion

96 Views Asked by At

In this video CLICK HERE, the professor says at 51:19 that if we add n Gaussian random variables and we see that sum further and further from our eyes, the result will look like a Brownian motion.

I do not not really understand what he means. I know that adding Gaussian random variables will be Gaussian and even if they are not originally Gaussian, the result will be Gaussian if n is high enough (by the central limit theorem). So I can not see any explanation on the internet that says that if we look further at the sum of Gaussian random variables, the result would look like a Brownian motion. Is there any interesting link that you can share with me concerning this fact?

Any help will be very appreciated!

1

There are 1 best solutions below

4
On BEST ANSWER

The word "add" (or "sum") is being used imprecisely here. You can construct an approximation to a Brownian motion by starting with $n$ independent standard Gaussian variables $X_1,\ldots,X_n$. Define the partial sums $$ S_0:=0,\quad S_1:=X_1, \quad S_2:=X_1+X_2,\ \ldots,\ S_k:=\sum_{i=1}^kX_i,\ \ldots,\ S_n:=\sum_{i=1}^nX_i. $$ Then plot $S_k$ versus $k$ in the plane, and connect the dots. This creates a continuous function. Under suitable scaling of the $x$ and $y$ axes, this function converges to a Brownian motion as $n\to\infty$. "Looking at this from farther and farther" is an informal way of saying "under suitable scaling".

There are many ways to construct Brownian motion: see Reference for the Construction of Brownian Motion. Donsker's Theorem, mentioned in the video, justifies the construction of Brownian Motion as the scaled limit of a random walk.