You might want to be more explicit about exactly what characteristics you want in this distribution,
That's hard. Mostly because I definitely do not want any formally defined distribution. Almost exactly the opposite in fact.
The problem with "random" is that on average, the points will be evenly distributed over the range (2D plane in this case). That's almost the exact definition of PRNG.
Whilst with enough iterations, all possible distributions -- including those where a majority of the points in the sample size tend to be grouped or clustered on one side or in one corner of the plane -- with even a relatively small plane, (500,500) and sample size 100 -- there are so many 'roughly even' distributions and so few 'lopsided' distributions, that I'd need to run billions of sets to ensure I'd tested a few of the lopsided ones. That's not practical.
So, I'm looking for a way to induce lopsided -- which pretty much means not 'roughly evenly distributed' -- distributions, without prescribing where or why the concentrated and sparse bits will be.
I can't think of any better description than: I want lopsided distributions that are completely randomly generated.
Not good I know, but its the best I've got.
In reply to Re^2: Randomly biased, random numbers.
by BrowserUk
in thread Randomly biased, random numbers.
by BrowserUk
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |