Understanding the normalization of a Gaussian

520 Views Asked by At

I have a Gaussian defined as follows:

$W(\theta) = j * exp(-0.5 * \theta^2 / \sigma^2)$.

I want to set $j$ such that $\frac{1}{360}\int_{-180}^{180} W(\theta)d\theta = 1$.

I'm using two values for $\sigma$: 36$^\circ$ and 72$^\circ$.

When I solve for $j$ setting $\sigma=36^\circ$, I find that

$j = 3.99 = 360/(\sigma\sqrt{2\pi})$.

However, when I solve for $j$ setting $\sigma=72^\circ$, I get

$j = 2.02,$

which is slightly greater than $1.99 = 360 / (\sigma\sqrt{2\pi})$.

Is there a single expression, like $360 / (\sigma\sqrt{2\pi})$, that captures $j$'s value across all $\sigma$?

EDIT:

I've made some progress on this, but I'm still not satisfied. I've got

$j = 360 / (\sigma\sqrt{2\pi}erf(127.279/\sigma))$,

where $erf()$ is the error function. I need a symbolic representation for 127.279.

1

There are 1 best solutions below

0
On BEST ANSWER

I've got it!

$j = 360 / (\sigma \sqrt{2 \pi} erf(\frac{180}{\sigma\sqrt{2}}))$.

(Not quite a "symbolic" representation, but I've gotten rid of that pesky -- read, harbinger of imprecision -- decimal point.)