3
$\begingroup$

Z is a standard normal and we have to prove that

\begin{equation} P(|z| \geq s) \leq {\sqrt{\frac{2}{\pi}}} {\frac{e^{\frac{-s^{2}}{2}}}{s}}\end{equation}

As far as I understand I have to use the formula for distribution function of a standard normal random variable, but I do not know how to deal with the ≥ inequality in the probability function as I have only seen examples with the ≤ inequality.

$\endgroup$
  • 1
    $\begingroup$ "I do not know how to deal with the ≥ inequality": Well, you can always use $P(X \geq z) = 1-P(X < z)$... $\endgroup$ – Florian Apr 25 '17 at 8:20
  • $\begingroup$ Consider the complement event: $P(|z| \ge s) = 1- P(|z|< s)$. $\endgroup$ – Crostul Apr 25 '17 at 8:21
  • $\begingroup$ @Florian Sorry but, what for? $\endgroup$ – Did Apr 25 '17 at 8:59
  • $\begingroup$ @Crostul Sorry but, what for? $\endgroup$ – Did Apr 25 '17 at 8:59
  • $\begingroup$ @Did: well you had asked how to convert $\geq$ into $\leq$, this is how. But as you see in the answer by Mau314, this was not even necessary to begin with. $\endgroup$ – Florian Apr 26 '17 at 8:39
1
$\begingroup$

I suggest to do it without considering the complement; you can prove this by studying explicitly the corresponding integral: For $s\geq 0$, we have $$P(|Z|\geq s)=2\int_s^\infty \frac{1}{\sqrt{2\pi}}\exp\left(-\frac{x^2}{2}\right)\mathrm{d}x.$$ Now via transformation you can artificially introduce a factor that corresponds to $\frac{1}{x}$. This factor can be bounded by $\frac{1}{x}\leq\frac{1}{s}$ and you can integrate the remaining $\exp(\cdot)-$term as usual.

In more detail, choose the transformation $z:=\frac{x^2}{2}$. Then the right hand side above is equal to $$\frac{\sqrt{2}}{\sqrt{\pi}}\int_{s^2/2}^\infty \frac{1}{\sqrt{2 z}}\exp(-z)\mathrm{d}z.$$ Now, using $\frac{1}{\sqrt{2z}}\leq\frac{1}{s}$ we obtain $$P(|Z|\geq s)\leq \frac{\sqrt{2}}{\sqrt{\pi}}\frac{1}{s}\int_{s^2/2}^\infty \exp(-z)\mathrm{d}z=\frac{\sqrt{2}}{\sqrt{\pi}}\frac{1}{s}\exp\left(-\frac{s^2}{2}\right).$$

$\endgroup$

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.