I have a question concerning this task:

Let $X$ be a random variable and $X_n=X+Y_n$ where $$E[Y_n]=\frac{1}{n}\quad\text{ and }\quad\operatorname{Var}(Y_n)=\frac{\sigma^2}{n}\quad \text{where }\sigma>0$$ Show that $X_n\xrightarrow{P}X$

Now I know that I can prove this by using the Chebyshev inequality, but can I also do it with the Markov inequality, meaning is this correct:

$$\lim\limits_{n\to \infty}P(|X_n-X|\geq \epsilon)=\lim\limits_{n\to \infty}P(|X+Y_n-X|\geq \epsilon)$$ $$=\lim\limits_{n\to \infty}P(|Y_n|\geq \epsilon)\leq \lim\limits_{n\to \infty}\frac{E[|Y_n|]}{\epsilon}= \lim\limits_{n\to \infty} \frac{1}{n\epsilon}=0$$

Here I came across $X-X$ and since both are random variables, I don't know if I can just cancel them? Is this correct?


The expected value of $Y_n$ is equal to $1/n$ ($\operatorname EY_n=1/n$), not the first absolute moment. We cannot say that $\operatorname EY_n=\operatorname E|Y_n|$. Hence, the last step in your proof is not justified.

However, we can use Lyapunov's inequality (which follows from Jensen's inequality) to conclude the proof. We have that $$ \operatorname E|Y_n|\le(\operatorname EY_n^2)^{1/2}=(\operatorname {Var}Y_n+(\operatorname EY_n)^2)^{1/2}=\biggl(\frac{\sigma^2}n+\frac1{n^2}\biggr)^{1/2}. $$

  • $\begingroup$ Thank you. But what about $X-X=0$, can I just subtract two random variables like that or is this not correct? $\endgroup$ – MarcE Jul 6 '16 at 15:21
  • 1
    $\begingroup$ @MarcE You're welcome! Yes, we can subtract two random variables like that. Strictly speaking, a random variable is a measurable function $X:\Omega\to\mathbb R$, where $(\Omega,\mathcal F,P)$ is the probability space. The difference of two random variables $X$ and $Y$ defined on the same probability space is given by $X(\omega)-Y(\omega)$ for each $\omega\in\Omega$ (let us observe that $X(\omega)$ and $Y(\omega)$ are just real numbers). If $X(\omega)=Y(\omega)$, then $X(\omega)-Y(\omega)=0$ for each $\omega\in\Omega$ which is what we use in the proof. $\endgroup$ – Cm7F7Bb Jul 6 '16 at 15:39

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.