# Continuous random variable and median problem

The definition of median, $$m$$, for a continuous random variable $$X$$ is $$P(X\leq m)=P(X\geq m)=\int_{-\infty}^m f(x)\text{d}x=\int_{m}^{\infty}f(x)\text{d}x=\frac{1}{2}$$

where $$f(x)$$ is the pdf of $$X$$.

Now the problem goes as follows:

Show that if $$X$$ is a continuous random variable, then $$min_{a} (E|X-a|)=E|X-m|$$

The work I have done for this problem:

WORK

Have that $$E|X-a|=\int_{\mathbb{R}} |x-a|f(x)\text{d}x=\int_{a}^{\infty}(x-a)f(x)\text{d}x-\int_{-\infty}^a (x-a)f(x)\text{d}x$$ Differentiation with respect to $$x$$ gives us $$\frac{d}{dx}E|X-a|=\int_{a}^{\infty} f(x)\text{d}x-\int_{-\infty}^a f(x)\text{d}x$$ Setting this equal to 0, we find that $$\int_{a}^{\infty} f(x)\text{d}x= \int_{-\infty}^a f(x)\text{d}x$$ We know that $$\int_{\mathbb{R}}f(x)\text{d}x=1$$ $$\implies \int_{a}^{\infty} f(x)\text{d}x= \int_{-\infty}^a f(x)\text{d}x=\frac{1}{2}$$ Hence, $$a$$ must be our median, by definition.

What I am confused on is showing the minimal property. I looked up what I could, and it appears that using the second derivative to show that we produce a positive value for f(a) present us with minimality, but I don't really understand why. Potentially a very naïve question, but why does f(a)>0 for second-derivative have significance? Is it because of the absolute-value property on $$X-a$$?

• self-study flag missing? Feb 10, 2015 at 7:18
• The derivative is not quite right: its sign is in error. Based on that, you should conclude that the median maximizes $E|X-a|$.
– whuber
Feb 10, 2015 at 15:58
• The variable of interest is $a.$ So the derivative should be taken with respect to $a.$ Feb 10, 2015 at 18:48

Based on your work, $$I(a) = \frac{d}{da} E|X-a| = \int_a^{\infty} f(x) ~ dx - \int_{-\infty}^a f(x) ~ dx$$ If $a=m$, then $I(m) = 0$.
Think of $I(a)$ geometrically. You have a probability mass functions $f(x)$. Geometrically $I(a)$ is the area of the right-half of the curve minus the left-half of the curve at the point $a$. Therefore, if $I(m) = 0$ and $a>m$ then, as you are weighing the curve now more to the left, $I(a) < 0$. By a similar argument, $I(a) > 0$ if $a<m$.
Ignore the second-derivative test. Use the first-derivative instead here. As $I(a) = 0$ at $a=m$, $I > 0$ for $a>m$, and $I<0$ for $a<m$, it follows that at $a=m$ we have a minimum point.
• Your argument would be correct if you used the correct derivative. With your expression for $I(a)$, it approaches $-1$ as $a\to\infty$ and $1$ as $a\to -\infty$, proving that $m$ is a maximum of $I$, not a minimum.