I am trying to find some reference with respect to the following problem: Given $n$ iid positive random variables $(X_i)_{i\leq n}$ of mean $E(X)>0$ and variance $V(X)$.

I wanted to know if there were some properties on the random variable: $Z_n = \frac{\sum (X_i -E(X))}{\sum X_i}$. Specifically I am interested in its convergence speed, something like the law of large numbers or Chebychev/Chernoff bounds. If it does not involve $V(X)$ it's even better.

Anyone knows where to look? Thanks


$$ Z_n = \frac{\sum_1^n (X_i -E(X))}{\sum_1^n X_i} = 1 - \frac{\mu}{\sum_1^n X_i/n}, $$ where $\mu = E(X)$.

Then, $$ Z = \lim_\limits{n \rightarrow \infty} Z_n =_d 1 - \frac{\mu}{G_n }, $$

where the limit is in distribution, with r.v. $G_n =_d N( \mu, \sigma^2/n)$

This convergence in distribution implies directly: $$ P(|G_n - \mu| >\delta ) = 2\Phi( -\frac{\delta \sqrt{n}}{\sigma}) $$

Convergence in distribution to a constant implies convergence in probability (to the same constant), that is, $G_n \rightarrow_P \mu$.

Then, $Z_n$ converges to zero with rate $1/n$ for a fixed error (as can be seen from the simple Wiki proof of weak LLN using Chebyshev inequality).

  • $\begingroup$ Thanks, but this is just a rewrite of the same equation. Can we say something then about the convergence speed of this new RV? $\endgroup$ – Gopi Mar 23 at 20:50

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.