Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Almost equal random variables

  1. Aug 4, 2009 #1
    I've been trying to solve the following question: Let X be a random variable s.t. Pr[|X|<+\infty]=1. Then for every epsilon>0 there exists a bounded random variable Y such that P[X\neq Y]<epsilon.

    The ideia here would be to find a set of epsilon measure so Y would be different than X in that set. However, it is not clear even that such a measurable set exists...

    Any help?
  2. jcsd
  3. Aug 4, 2009 #2
    To show there is some interval (-a, a) such that [tex]\int_{-a}^a f(x) dx > 1-\epsilon[/tex], apply the definition of an improper integral to [tex]\int_{-\infty}^\infty f(x) dx = 1[/tex]. Have Y=X over (-a,a).
  4. Aug 5, 2009 #3
    For a large positive number [itex]\lambda[/itex], define [itex]Y = X[/itex] on the set [itex]\{|X| \le \lambda\}[/itex] and [itex]Y=0[/itex] on the remaining part of the sample space. Now all you have to do is choose [itex]\lambda[/itex] large enough to get the conclusion you want.
  5. Aug 5, 2009 #4
    It would remain to show that such $\lambda$ exists. We can definte the sets A_n=\{\omega\in \Omega: |X(\omega)|<n\}. Since A_n is measurable, as X is a random variable, and A_n \uparrow \{\omega: |X|<+\infty}, we have that \lim P(A_n) = P(\{\omega: |X|<+\infty}. Therefore, for all \epsilon, there is $m$ large enough so that P(A_n)>1-\epsilon. Then we can define Y=X in A_m and Y=0 otherwise.

Share this great discussion with others via Reddit, Google+, Twitter, or Facebook