1
$\begingroup$

I need to show that if $X_n \rightarrow X$ and $X_n \rightarrow Y$, then $X\overset{\text{a.s.}}{=}Y$ for convergence in probability, convergence almost surely, as well as for convergence in mean and quadratic mean ($\mathcal L^1$ and $\mathcal L^2$ convergence).

Convergence in Probability:

For any $\epsilon>0$ and for any $n\in\mathbb N$ we have

$$\begin{align} \mathbb P(|X-Y|\geq\epsilon) &\leq\mathbb P(|X-X_n|+|X_n-Y|\geq\epsilon)\\\\ &\leq\mathbb P\left((|X-X_n|\geq\epsilon/2)\cup(|X_n-Y|\geq\epsilon/2)\right)\\\\ &\leq\mathbb P(|X-X_n|\geq\epsilon/2)+\mathbb P(|X_n-Y|\geq\epsilon/2) \end{align}$$

so that

$$\mathbb P(|X-Y|\geq\epsilon)\leq\lim_{n\rightarrow\infty}\mathbb P(|X-X_n|\geq\epsilon/2)+\mathbb P(|X_n-Y|\geq\epsilon/2)=0$$

Since $$\{|X-Y|>0\}=\underbrace{\bigcup_{n=1}^\infty \underbrace{\left\{|X-Y|>\frac{1}{n}\right\}}_{=\emptyset}}_{=\emptyset}=\emptyset$$

we have that $\mathbb P\{|X-Y|>0\}=0$ and so $\mathbb P(X\ne Y)=0$. Hence $\mathbb P(X= Y)=1$ which means that $X\overset{\text{a.s.}}{=}Y$.

Convergence Almost Surely:

Since almost sure convergence implies convergence in probability, the result follows immediately from the last part. However, I'd like to show this without making use of that result. Since $X_n$ converges almost surely to both $X$ and $Y$ then $\mathbb P(\lim_{n\rightarrow\infty}X_n=X)=1$ and $\mathbb P(\lim_{n\rightarrow\infty}X_n=Y)=1$. From here it seems obvious to me that $X\overset{\text{a.s.}}{=}Y$ but I'm not sure how to show this formally.

Convergence in Mean:

$$\begin{align} \mathbb E(|X-Y|) &\leq\mathbb E\left(|X-X_n|+|X_n-Y|\right)\\\\ &=\mathbb E\left(|X-X_n|)+\mathbb E(|X_n-Y|\right) \end{align}$$

so

$$\mathbb E(|X-Y|)\leq\lim_{n\rightarrow\infty}\mathbb E(|X-X_n|)+\mathbb E(|X_n-Y|)=0$$

so $X\overset{\text{a.s.}}{=}Y$

Convergence in Quadratic Mean:

I tried continuing with the same logic but it's not the case that

$$ \mathbb E(|X-Y|^2)\leq\mathbb E\left(|X-X_n|^2+|X_n-Y|^2\right)$$

so I'm not sure how to proceed.

Is my reasoning correct for the first and third? How can I proceed with the other two?

$\endgroup$
1
  • 1
    $\begingroup$ Yes, first and third seems correct. For the second, note that $X_n \to X, X_n \to Y$ almost surely, means you have a set $A$ of full measure, such that for $\omega \in A$ you have $X_n(\omega) \to X(\omega)$ and $X_n(\omega) \to Y(\omega)$ as a sequence of numbers, so easily $|X(\omega)-Y(\omega)| \le |X(\omega)-X_n(\omega)| + |Y(\omega) - X_n(\omega)| \to 0$ which means that you have $X=Y$ on $A$, hence almosts surely. For the fourth: Minkowsky Inequality and goes the same as third one $\endgroup$ Oct 4 '20 at 19:58
0
$\begingroup$

I think I can answer my own question now for convergence in quadratic mean. By Minkowsky,

$$\begin{align*} \left(\mathbb E\left(|X-Y|^2\right)\right)^{1/2} &=\left(\mathbb E\left(|(X-X_n)+(X_n-Y)|^2\right)\right)^{1/2}\\\\ &\leq \left(\mathbb E\left(|X-X_n|^2\right)\right)^{1/2}+\left(\mathbb E\left(|X_n-Y|^2\right)\right)^{1/2} \end{align*}$$

so

$$\begin{align*} \left(\mathbb E\left(|X-Y|^2\right)\right)^{1/2} &\leq\lim_{n\rightarrow\infty} \left(\mathbb E\left(|X-X_n|^2\right)\right)^{1/2}+\left(\mathbb E\left(|X_n-Y|^2\right)\right)^{1/2}\\\\ &=0 \end{align*}$$

so $X\overset{\text{a.s.}}{=}Y$

$\endgroup$
0
$\begingroup$

It follows from convergence a.s. and from convergence in $L_p$ that there is convergence in probability. So it was enough to condsider convergence in probability.

$\endgroup$

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.