1

How can I show that if $X_n \overset{p}{\to} X$ and $Y_n \overset{p}{\to} Y$, then $(X_n, Y_n) \overset{p}{\to} (X,Y)$? that is, if $X_n$ converges in probability to $X$, and $Y_n$ converges in probability to $Y$, then we have convergence in the joint sense. I've tried to use an epsilon approach, but cannot seem to get a bound. Does anyone have any hints? Thanks!

user136503
  • 2,359
  • 16
  • 26

1 Answers1

2

For any $\epsilon>0$

$$P\left\{d\left((X_n,Y_n),(X,Y)\right)>\epsilon\right\}\le P\{|X_n-X|>\epsilon/\sqrt{2}\}+P\{|Y_n-Y|>\epsilon/\sqrt{2}\}\to 0.$$

Here, $d(z, w)=\|z-w\|_2$.

  • Could I ask where you got the $\sqrt{2}$ in the epsilons? – user136503 Nov 23 '15 at 04:52
  • @user136503 $$P\left{d\left((X_n,Y_n),(X,Y)\right)>\epsilon\right}=P{(X_n-X)^2+(Y_n-Y)^2>\epsilon^2}\le \cdots$$ –  Nov 23 '15 at 05:01
  • Thank you! I saw a reference to this problem being done with a supremum norm, is that what you had here in the dots or is the supremum norm just a more general bound? thanks! – user136503 Nov 23 '15 at 05:07
  • Nope. Dots are for the second term in the inequality: $$P{A^2+B^2>\epsilon^2}\le P{A^2>\epsilon^2/2}+P{B^2>\epsilon^2/2}$$ –  Nov 23 '15 at 05:12
  • Got it! Thanks so much!! – user136503 Nov 23 '15 at 05:17