Let $(X_n)_{n\geq 1}$ be i.i.d. random variables with uniform distribution on the interval $(0,1)$.

I need to prove that the following sequence of random variables $(Y_n)_n$ defined by:

$$Y_n = \frac{X_1^2+\dots+X_n^2}{X_1+\dots+X_n}$$

converges almost surely, and then compute

$$\lim_{n\to\infty} \mathbb{E}(Y_n)$$

It does not seem to be a very difficult problem, but I am stuck.

Thanks a lot!


To prove the almost sure convergence write

$$Y_n = \frac{X_1^2+\ldots+X_n^2}{n} \frac{n}{X_1+\ldots+X_n}$$

and apply the strong law of large numbers (twice).

For the second part of the problem, use the fact that $0 \leq X_i \leq 1$ to show that

$$0 \leq Y_n \leq 1.$$

Combine the dominated convergence theorem with the first part of the problem to compute the limit $\lim_n \mathbb{E}(Y_n)$.

  • $\begingroup$ Thanks a lot!!! Very nice and helpful answer! $\endgroup$ – Jaime Arboleda Castilla May 6 '18 at 21:03
  • $\begingroup$ @JaimeArboledaCastilla You are welcome. $\endgroup$ – saz May 7 '18 at 7:17

Hint: you can use the Strong Law of Large Numbers (SLLN) just write

$$ Y_n = \frac{X_1^2 + ... + X_n^2}{n} \frac{1}{ \frac{X_1+...+X_n}{n}}. $$ Now, both fractions converge a.s. in view of the SLLN, and hence so does the $Y_n$. You also get the expectation using in addition the boundedness of the sequences.

  • 1
    $\begingroup$ regarding the last part: is that really "easily" a consequence? The first part shows that $Y_n$ converges a.s., but you need a bit more to guarantee convergence in mean as well. $\endgroup$ – Clement C. May 5 '18 at 18:28
  • $\begingroup$ fair enough. Not a consequence of a.s. convergence alone, but the boundedness of the sequence as well, plus a bit of justification on top of those. Thanks, I'll reword that. $\endgroup$ – Hayk May 5 '18 at 18:38

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.