I intend to find the Gaussian function, where $p$ is the probability, area under its curve, within the boundary $r_{min}=- r_0$ and $r_{max} = +r_0$.

Consider the following Gaussian function:

$$ f(r) = \frac{1}{\sigma \sqrt{2\pi}}e^{-\frac{1}{2}\frac{r^2}{\sigma^2}}$$

To solve this, I employed the ff:

$$p =\int_{r_{min}}^{r_{max}} \frac{1}{\sigma \sqrt{2\pi}}e^{-\frac{1}{2}\frac{r^2}{\sigma^2}}dr$$

But: $$\int e^{-\alpha r^2} dr = \frac{\sqrt{\pi}erf(\sqrt{\alpha}r)}{2\sqrt{\alpha}}$$

Then: $$p=erf(\sigma \sqrt{2}r_0)$$

So: $$\sigma=erf^{-1}(p)/(\sqrt{2}r_0)$$


$$ f(r) = \frac{1}{erf^{-1}(p)/(\sqrt{2}r_0) \sqrt{2\pi}}e^{-\frac{1}{2}\frac{r^2}{(erf^{-1}(p)/(\sqrt{2}r_0))^2}}$$

The solution seems sound, but when I graphed it, I dont see that the area under the curve within $r_0$ is actually $p$.

Below are the relevant images of what I am talking about.

The top image is the Gaussian graph I obtained while the bottom image is what I am expecting. The red line is where $r_0$ is.

The plot shown was only half of the curve, so I am expecting to get $p/2$ of the area

enter image description here

  • $\begingroup$ You made two mistakes: (1) scaling between Erf and Normal density (2) algebra when inverting Erf (division done wrong). $\endgroup$ – Lee David Chung Lin Feb 14 at 14:14
  • $\begingroup$ I corrected the algebra, its just typo. About the scaling? What would be the correct way for that? I am not that familiar with scaling in erf and normal density $\endgroup$ – Jones G Feb 14 at 14:23

The definition of the error function is $$\mathrm{Erf}(x) = \int\limits_{t = -x}^{t=x} \frac1{\sqrt\pi}e^{-t^2} \,dt$$

With that in mind, we have \begin{align} p &= \int\limits_{r = -r_0}^{r=r_0} \frac{1}{\sigma \sqrt{2\pi}}e^{-\frac{1}{2}\frac{r^2}{\sigma^2}}\,dr \qquad \text{, let}~~ t = \frac{r}{\sigma \sqrt 2} \\ &= \int\limits_{r = -r_0}^{r=r_0} \frac1{\sqrt\pi} e^{-( \frac{r}{\sigma \sqrt 2})^2 } \frac{dr}{\sigma \sqrt 2} \\ &= \int\limits_{t = -t_0}^{t=t_0} \frac1{\sqrt\pi}e^{-t^2} \,dt \qquad \text{, note that}~~ t_0 = \frac{r_0}{\sigma \sqrt 2} \\ &= \mathrm{Erf}\Bigl( \frac{r_0}{\sigma \sqrt 2} \Bigr) \end{align} $$ \implies \mathrm{Erf}^{-1}(p) = \frac{r_0}{\sigma \sqrt 2} \qquad \implies \sigma = \frac{r_0}{\sqrt 2 \mathrm{Erf}^{-1}(p)}$$ Then you can proceed to carry out your further plans.

  • $\begingroup$ I got what I expected! $\endgroup$ – Jones G Feb 14 at 14:28

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.