1
$\begingroup$

As we know, if a sequence of distribution functions $\{F_n\}$converge weakly to $F$, i.e. $\lim_{n \to \infty}F_n(x)=F(x)$ for all continuous points $x\in\mathbb{R}$ of $F$, then for the corresponding quantile functions, we have $\lim_{n\to\infty}Q_n(u)=Q(u)$ for all continuous points $u\in(0,1)$ of Q.

So my question is if $u$ is not a continuous point of Q, can we find an example that $Q_n(x)$ does not converge?

The discontinuous points (jumps) in $Q$ corresponds to flat points in $F$, but I don't know how to construct such a $\{F_n\}$.

$\endgroup$
1
$\begingroup$

Let $F_n(x)=0$ for $x\le 0$, $F_n(x)=1$ for $x\ge 3$, $F_n(x)=\frac12+\frac{(-1)^n}{n}$ for $1\le x\le 2$. Extend $F_n$ to the intervals $(0,1)$ and $(1,2)$, for example affinely: as $F(x)=F(1)x$ for $0<x<1$ and $F_n(x)=(1-F_n(2))(x-2)+1$ for $2<x<3$.

As $n\to \infty$, the sequence $F_n$ converges uniformly. The median of the random variable described by $F_n$ is less than $1$ when $n$ is even, and greater than $2$ when $n$ is odd. Hence, the median does not have a limit as $n \to\infty$.

$\endgroup$

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.