Is it possible to set up a probability space with sample space $\Omega=\{1,2,\dots\}$ to model a 'uniformly chosen positive integer'?

343 Views Asked by At

$\textbf{The Problem:}$ Is it possible to set up a probability space with sample space $\Omega=\{1,2,\dots\}$ to model a 'uniformly chosen positive integer'?


$\textbf{My Thoughts:}$ Suppose that we can indeed set up such a probability space $(\Omega,\mathcal F,P)$. Observing that outcomes must be equally likely due to the uniformly chosen hypothesis, we may assume that there exists some $0<\varepsilon<1$ such that for all $n\in\Omega$ we have that $P(\{n\})=\varepsilon$. Then since $\bigcup_{n\in\mathbb N}\{n\}=\Omega$ is a union of pairwise disjoint sets, countable additivity implies that $$P(\Omega)=P\left(\bigcup_{n\in\mathbb N}\{n\}\right)=\sum_{n=1}^\infty\varepsilon=\infty>1,$$ and we have a contradiction. Therefore, no such probability space can be set up.


Could anyone please provide some feedback on the proof above?

Any comments are much appreciated, and thank you very much for your time.

1

There are 1 best solutions below

0
On BEST ANSWER

Answering for the sake of being able to resolve the question, but making this a community wiki since I'm not contributing anything of substance:

You've got it!