0
$\begingroup$

Simple question, but every time I think I have the answer, I get stuck.

Let $ F(x)=\frac{1}{x^2}*(2*\Phi(x)-1) +(2/3)*\frac{1}{x}*(\phi(x)) +(2/3)*\frac{4}{x^3}*(\phi(x)-\phi(0))$

where $\phi$ is the standard normal PDF and $\Phi$ is the standard normal CDF.

Show that $F(x)>0$ for all $x>0$

So far I have that the limit as x->0 = $\infty$ and that as x->$\infty$, $ F(x)->0$

But, then showing the first derivative is negative leads to the same problems as showing the equation above is positive and I'm left where I started.

$\endgroup$
0
$\begingroup$

Here is one possible approach.

Assuming that you mean you want to prove that $$F(x) = \left.\left.\frac{1}{x^2}\right[2\Phi(x)-1\right] + \frac{2}{3x}\phi(x) + \left.\left.\frac{8}{3x^3}\right[\phi(x)-\phi(0)\right] > 0 ~~\text{for}~ x > 0,$$ which holds if $$(2\Phi(x)-1) + \frac{2}{3}x\phi(x) + \left.\left.\frac{8}{3x}\right[\phi(x)-\phi(0)\right] > 0 ~~\text{for}~ x > 0,$$ notice that $\Phi(x)-\Phi(0) = \Phi(x)-\frac{1}{2}$ is the probability that a standard normal random variable has value in $[0,x]$ and that $\frac{\mathrm d}{\mathrm dx}\phi(x) = \phi^{\prime}(x) = -x\phi(x)$. Thus, we can write that for $x >0$, $$\begin{align*} x^2F(x) &= (2\Phi(x)-1) + \frac{2}{3}x\phi(x) + \left.\left.\frac{8}{3x}\right[\phi(x)-\phi(0)\right]\\ &= 2\int_0^x \phi(t)\,\mathrm dt + \frac{2}{3}\phi(x)\int_0^x 1\,\mathrm dt + \frac{8}{3x} \int_0^x \phi^{\prime}(t)\,\mathrm dt\\ &= \int_0^x \left[2 - \frac{8t}{3x}\right]\phi(t) + \frac{2}{3}\phi(x)\,\mathrm dt \end{align*}$$ The integrand has value $2\phi(0) + \frac{2}{3}\phi(x) > 0$ at $t = 0$, $\frac{2}{3}\phi(x)$ at $t = \frac{3}{4}x$, and value $0$ at $t = x$. So if the integrand can be proved to be nonnegative on $(0,x)$, you are done.

$\endgroup$

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.