I'm working throught Wasserman's "All of Statistics" book. When proving convergence of random variables/distributions in chapter 5, he lists the following inequality:

$$F_n(x) = \mathbb{P}(X_n\le x)=\mathbb{P}(X_n\le x, X\le x+\epsilon) + \mathbb{P}(X_n\le x, X\gt x+\epsilon)\\ \le \mathbb{P}(X \le x + \epsilon) + \mathbb{P}(\vert X_n - X \vert \gt \epsilon)\\ = F(x + \epsilon) + \mathbb{P}(\vert X_n - X \vert \gt \epsilon) $$

Where $\mathbb{P}$ is the probability that the random and $F$ is the cdf of the random variable $X$

Here, $X$ and $X_n$ are different random variables (it's part of a proof showing $X_n$ converges to $X$).

What I don't understand is how he arrives at the inequality on the second line. Could someone help me understand this. I feel like my lack of some essential fundamental knowledge of statistics is being exposed by not being able to figure this out.


  • 1
    $\begingroup$ If two numbers are separated by at least $\epsilon,$ then the absolute value of their difference is at least $\epsilon.$ $\endgroup$
    – whuber
    Oct 25 '18 at 23:34

Note that $X_n \le x$ and $X \le x+\epsilon$ implies $X \le x+ \epsilon$, that is

$$\{\omega \in \Omega| X_n (\omega) \le x, X(\omega) \le x+ \epsilon \} \subseteq \{\omega \in \Omega| X(\omega) \le x+ \epsilon \} $$

Hence we have $\mathbb{P}(X_n \le x, X \le x+\epsilon)\le \mathbb{P}(X \le x+\epsilon).$

Also if $$X_n \le x, X > x+\epsilon$$ then we have

$$X_n \le x \le x+\epsilon < X$$

and $|X_n - X|=X-X_n > x+\epsilon - x=\epsilon.$

That is $$\{\omega \in \Omega| X_n (\omega) \le x, X(\omega) > x+ \epsilon \} \subseteq \{\omega \in \Omega| X_n(\omega)-X(\omega) > \epsilon \} $$

Hence $\mathbb{P}(X_n \le x, X > x+\epsilon) \le \mathbb{P}(X_n -X > \epsilon).$

Summing up the two inequalities would give us the result.


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.