Random walk - why is the STD equal sqrt(n)

AI Thread Summary
The discussion centers on the properties of a typical random walk, where each step is an independent indicator variable taking values of 1 or -1 with equal probability. The initial confusion arises from incorrectly defining the sum of steps, leading to a miscalculation of variance. The correct formulation shows that the sum of the steps, denoted as Sn, equals the sum of individual steps rather than n times a single step. Consequently, the expectation of Sn is zero, and the variance is n, leading to the conclusion that the standard deviation is √n. The clarification emphasizes the importance of accurately defining the random variables involved in the calculations.
gony rosenman
Messages
11
Reaction score
4
Member warned that the homework template must be used
typical random walk :
one step forward or backward with equal probability and independence of each step , what is the expectation and Variance .

so i define indicator variable xi ={1 or -1 with equal probabilty .
E(xi) = 0
Var(xi) = 1

now define Sn as the sum of i=1,...,n
each step is independent and identical to the other so
i can say that Sn = nxi
now from definition of Expectation and Variance :
E(Sn) = E(nxi) = nE(xi) = 0
Var(Sn) = Var(nxi) = n2Var(xi) = n2 ⇒σsn=n
but i know for certain that actually the answer is σsn=sqrt(n)
how is that ?
please
thank you!
 
Physics news on Phys.org
##S_n## is the sum of ##x_i: i=1..n##
That is not ##nx_i##.

##S_n## would be {##-n, ... , +n##} with a binomial distribution.
 
Last edited:
.Scott said:
##S_n## is the sum of ##x_i: i=1..n##
That is not ##nx_i##.

##S_n## would be {##-n, ... , 0, ..., +n##} with a binomial distribution.
could you elaborate further?
 
gony rosenman said:
could you elaborate further?
For n=4:
##Sn## would be {-4, p=1/16; -2, p=4/16; 0, p=6/16; +2, p=4/16; +4, p=1/16}
or {-4, -2, -2, -2, -2, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 4}

Actually, in post #2, that zero I put in the middle is wrong (I'll edit it):
For n=5:
##Sn## would be {-5, p=1/32; -3, p=5/32; -1, p=10/16; +1, p=10/32; +3, p=5/32; +5, p=1/32}
or {-5, -3, -3, -3, -3, -3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 5}
 
.Scott said:
For n=4:
##Sn## would be {-4, p=1/16; -2, p=4/16; 0, p=6/16; +2, p=4/16; +4, p=1/16}
or {-4, -2, -2, -2, -2, 0, 0, 0, 0, 0, 0, 2, 2, 2, 2, 4}

Actually, in post #2, that zero I put in the middle is wrong (I'll edit it):
For n=5:
##Sn## would be {-5, p=1/32; -3, p=5/32; -1, p=10/16; +1, p=10/32; +3, p=5/32; +5, p=1/32}
or {-5, -3, -3, -3, -3, -3, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 3, 3, 3, 3, 3, 5}
yes but how do you show that σ=√n
 
I was just showing that your logic was wrong. I wasn't computing what it would actually be.
 
.Scott said:
I was just showing that your logic was wrong. I wasn't computing what it would actually be.
are you sure you are capable to assist me ?
i think you might not know what an indicator random variable is and thus you confuse me and might cause actual damage , even though I'm sure your intentions are good
 
o.k so i figured it out by myself , sor anyone who might stumble upon this confusion i will explain :
so sn ≠ nxi but actually sn = ∑ni=1 xi
so E(sn) = E(x1) +...+ E(xn) = 0
and Var(sn) = Var(x1) +...Var(xn) = n ⇒σsn= √n
my mistake was for doing sn = nxi when actually it is sn = ∑ni=1 which is tremendously different .
good luck!
 
I am using the same nomenclature that you introduced.
if ##x_i## = {##-1, 1##} with equal distribution
then the distribution you will get with ##S_n## will be what I showed.
When I take the population standard deviation of some example ##S_n##'s, I get ##\sqrt{n}##.

What is very likely is that your ##nx_n## is not normal multiplication. But if it isn't, you would be running into trouble when you use it as normal multiplication later in your work.
 
  • Like
Likes gony rosenman
  • #10
.Scott said:
I am using the same nomenclature that you introduced.
if ##x_i## = {##-1, 1##} with equal distribution
then the distribution you will get with ##S_n## will be what I showed.
When I take the population standard deviation of some example ##S_n##'s, I get ##\sqrt{n}##.

What is very likely is that your ##nx_n## is not normal multiplication. But if it isn't, you would be running into trouble when you use it as normal multiplication later in your work.

What is more likely is that the OP originally made an error, and wrote down something without giving it a lot of thought.

In fact, the original argument of the OP is almost right: each ##X_i## has mean zero, so each has variance ##\text{Var}(X_i) = E(X_i^2)##. However, since it really is true that all the ##X_i^2 = 1## are identical, it follows that the variance is ##\sum 1 = n.##
 
Last edited:
  • Like
Likes .Scott
Back
Top