Convergence in distribution with finite mean

431 Views Asked by At

I'm preparing myself for the final exam of my graduate Probability Theory course and was stuck with another one of the exercises our professor gave us.

Let $X_n, n=1,2,\ldots,$ and $X$ be nonnegative random variables such that $X_n \rightarrow X$ in distribution as $n \rightarrow \infty.$

Prove or disprove: If $X_n, n=1,2,\ldots,$ and X have finite means $\mu_n,n=1,2,\ldots,$ and $\mu$, respectively, then $\mu_n \rightarrow \mu$ as $n \rightarrow \infty$.

Here is my current work:

$\mathbb{E}[X_n]=\mu_n < \infty, \mathbb{E}[X] = \mu < \infty.$ We want to show that $\lim_{n\rightarrow \infty}\mathbb{E}[X_n] = \mathbb{E}[X],$ i.e. $\mathbb{E}[X_n - X]=0.$

I think the statement is wrong, and made the following counterexample to disprove it:

Take $X \neq Y$ almost surely, and $X = Y$ in distribution. Then $X_n \rightarrow X = Y$ in distribution, but $\mathbb{P}(|X_n - Y|>\varepsilon) = \mathbb{P}(|X-Y| > \varepsilon) > 0.$

Is this correct?

1

There are 1 best solutions below

1
On BEST ANSWER

As @RodrigoRibeiro pointed out, your counterexample doesn't work because $X=Y$ in distribution implies $\mathbb{E}X= \mathbb{E}Y$ and therefore the sequence $\mu_n = \mathbb{E}X$ converges to $\mu = \mathbb{E}Y$.

Nevertheless, the statement is wrong. Consider $((0,1),\mathcal{B}((0,1)))$ endowed with the Lebesgue measure and

$$X_n(\omega) := n^2 1_{(0,1/n)}(\omega), \qquad \omega \in (0,1).$$

Then $X_n \to X:=0$ almost surely (hence, in particular, $X_n \to X$ in distribution), but

$$\mathbb{E}X_n = n$$

does not converge to $\mathbb{E}X=0$.