There is a statement which I am trying to evaluate to whether true or false:

Let $F_X(x)$ be a cumulative distribution function (cdf) of a r.v. $X$. Then the following $ G_X(x) $ is also a cdf: $$ G(x) = \frac{F_X(x) + 1 - F_X(-x)}{2} $$

It's very clear this is true for distributions which are symmetric around 0, because then $P(X\leq - x) = P(X \geq x)$. It's also simple to show that $ \lim_{x \to \infty} G(x) = 1 $ when $F_X$ is a distribution.

I tend to believe this is a true statement. What is bothering me is the continuity in the right it must have. So I tried to think of a counterexample. Let:

$$ F_X(x) =\begin{cases} 0, \text{ if } x<0 \\ \frac{1}{15} + \frac{2t}{3} \text{ if } x \in [0,1) \\ 1 \text { if } x\geq 1 \end{cases} $$

which is a mixture of a Bernoulli r.v. with $p=4/5$ and an uniform in $(0,1)$ with weights $1/3$ and $2/3$. Now I believe that for $x<0$, $G(x)$ is not continuous in the right (particularly in $x=-1$).

Does it suffice as a counterexample? Or is the statement right?



You counterexample is correct. And any $F$ that is discontinuous in at least one point will be a counterexample. For example, take $F(x) = \mathbb{I}_{x \geqslant 0}$. Then $G(0) = \frac{1 + 1 - 1}{2} = \frac{1}{2}$, but if $x > 0$ we have $G(x) = 1$, thus $\lim_{x \to 0+} G(x) = 1$.

And in general, $\lim_{x \to x_0+} G(x) = G(x_0) - \lim_{x \to x_0+} \frac{F(-x) - F(-x_0)}{2} = G(x_0) - \lim_{x \to -x_0+}\frac{F(x) - F(-x_0)}{2}$, so $G$ is continuous in right at $x_)$ iff $F$ is continuous in left in $-x_0$.

  • $\begingroup$ Excelent! Thanks! :) $\endgroup$ – M.Gonzalez Apr 30 '19 at 13:18

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.