0
$\begingroup$

I'm trying to show that if $X_n$ converges in probability to 0 and $Y_n$ converges in probability to 0, then $X_n+Y_n$ converges in probability to $0$, ie the sum rule for probability limits.

What I'm trying to do is to show it directly using $\epsilon, \delta$ arguments as opposed to appealing to the CMT, to improve my understanding of convergence. But I get stuck almost immediately because I don't know how I can use bounds for the probability each sequence exceeds some value to get a bound for the probability the sum exceeds some value...

Here's where I get stuck:

For every $\epsilon>0$ and every $\delta>0$ I have an $N$ s.t $n>N \implies P( |X_n|>\epsilon)<\delta$ and $P(|Y_n|>\epsilon)<\delta$. I also know that $|X_n+Y_n| \leq |X_n|+|Y_n|$ and that therefore $P( |X_n+Y_n|>\epsilon) \leq P( |X_n|+|Y_n| >\epsilon)$. Now it seems obvious that if $P( |X_n|>\epsilon)$ and $P(|Y_n|>\epsilon)$ can be made as small as we want, the RHS can be bounded, but I can't quite work out how. I can't seem to relate the probability of the sum to the probabilities I can bound.

$\endgroup$
2
$\begingroup$

You can argue like this. If $\left| X_n \right| + \left| Y_n \right| > \epsilon$, then it must be the case that $\left| X_n \right| > \frac{\epsilon}{2}$ or $\left| Y_n \right| > \frac{\epsilon}{2}$. Because if both $\left| X_n \right|, \left| Y_n \right| \leq \frac{\epsilon}{2}$, then:

$$ \left| X_n \right| + \left| Y_n \right| \leq \frac{\epsilon}{2} + \frac{\epsilon}{2} = \epsilon $$

So now you can use the union bound

$$ Pr( \left| X_n \right| + \left| Y_n \right| > \epsilon ) \leq Pr \left( \left| X_n \right| > \frac{\epsilon}{2} \lor \left| Y_n \right| > \frac{\epsilon}{2} \right) \leq Pr \left( \left| X_n \right| > \frac{\epsilon}{2} \right) + Pr \left( \left| Y_n \right| > \frac{\epsilon}{2} \right) $$

which you can bound in terms of $\delta$.

$\endgroup$
1
  • $\begingroup$ Thanks! Learned something new about sums today, namely that they imply this very useful disjunction...:) $\endgroup$ – CloseToC May 24 '15 at 19:04

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.