I'm working through an exercise related to a recent convergence/limit lecture for a probability theory course I'm taking and I seem to be doing something wrong, as I know the solution but seem to be getting off track at some point. The question:

Given an i.i.d. sequence $X_1,X_2...$ uniformly distributed on $[0,1]$, determine

$\lim_{n \to \infty} \sqrt[n]{X_1...X_n}$

With probability 1.

So, first I'm thinking that given the language "with probability 1" I should be thinking strong law of large numbers. So, with a sequence of uniform random variables on $[0,1]$, the variable converges a.s. to $\frac{1}{2}(a+b)$, or $\frac{1}{2}(0+1) = \frac{1}{2}$

Now, as n tends to infinity, the nth root of $\frac{1}{2}$ goes to 1.

The solution to the exercise is given as $e^{-1}$

Can anyone offer guidance on where I'm going wrong?

Thank you!


The problem with your approach is that the Strong Law of Large Numbers is a statement about the arithmetic mean of $X_1,\dots,X_n$, while this problem is considering the geometric mean.

However, note that $$ \log(\sqrt[n]{X_1\cdots X_n})=\frac{\log(X_1)+\dots+\log(X_n)}{n} $$ and since the random variables $\{\log(X_n)\}_n$ are i.i.d., it is enough to determine their expected value, then appeal to the Strong Law of Large Numbers.

  • $\begingroup$ Can't believe I didn't recognize that as a geo mean.. got it now thanks $\endgroup$ – Archetupon Nov 8 '16 at 23:57

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.