# Gaussian distribution centre

hello

anyone a idea to change the centre of the gaussian distribution (of the gaussian spread module).

sounds easy but…

?? i thought it was the first pin called “input”?
may be i’m wrong

ähm yes. but i only want to change the centre - not the borders.
andersgesagt: i need sth like a Chi-Quadrat-Verteilung
ok after transforming the centre to the left or the right it is not a gauss. but i did not know how to explain.

uh
ok
i’m keeping out of there.

well. here is a better explanation:

## gaussian spread module works like thathttp://www.rauhfaserrakete.de/verteilung1 var1 = spread amount var2 = standard deviation input

and i´m looking to have sth like that
http://www.rauhfaserrakete.de/verteilung2
var1 & var2 like the gaussian spread module
but also var4 for the possibility to change the position of the centre.

anyone a idea?
i tried to use sort,shift,scroll with the output of the gaussian spread - but it didnt work

?

Hi hrovac,

Try Edgemorph (Value) available at Sven’ page.

Cheers,
aR

and you can try a random spread, which has a rectangular distribution and multiply it with the curve you like…

hi!

another idea:

If you look at the second image (“Verteilung”, distribution), then you see that these all are functions which can be inverted (f^-1) (“Umkehrfunktion”).
When you have that, then on the x-axis are values between 0…1 (which can be easily generated by a random node) and on the y-axis you have your positive real numbers (the output of your gamma distribution).

A
You only need to somehow construct a patch, that acts like one of these inverse distribution functions.
So you need the inverse of the “Verteilungsfunktion” y = F(n,x). (something like x = F(-1)(n,y))…

B
if getting F(-1) is too hard, another way would be to somehow construct the original curve F(n,x) with a expr node and then create many value pairs and store them inside a big spread. just use a spread of x values and store the y=F(n,x) values which you get as a result of evaluating the expr. so you have a spread of points on that function curve (F).

Now, to come from a y value to a x value
(nearly the inverse func, table lookup + linear mapping):

After that you can use the interval node to reconstruct the x0 value to a newly generated y0 value (between 0 and 1, created by the random node) and get a pair of x values (interval x1…x2) in between those the corresponding point x0, y0 must be.
when you have these two points x1,y1…x2,y2 you can map somehow

y0 in y1…y2 -->> x0 in x1…x2 (map node)

if you go with that second approach (B) you will probably need to generate more points for x in 0…5 than for 5…10 so that for each y generated by random there will be a good (small) interval which maps it to an x value.

but we can discuss this another time?

ohoh big thx for all the ideas.

@gregsn: i´m not sure if i understand all you wrote. but i made what you said. i took a verteilungsfkt and played around with it.
now i have what i wanted(spreadable):

verteilung.v4p (17.5 kB)