RE: random numbers

[Ted Harding]

On 15-Feb-99 Gerd Bürger wrote:
> I am not that familiar with random number generators. But is it
> reasonable that for 2 independant random series, generated with
> 
> e=normal_rnd(100000,2);
> 
> I always find cor(e(:,1),e(:,2)) in the order of 1e-3?
> 
> For my taste, this is a little too much.
> 
> Is it?

For independent Normal X,Y, the standard deviation of the correlation
coefficient R is 1/sqrt(n) -- in this case, therefore, approx = 0.003
For large values of n (such as 100000), the distribution of R is close
to Normal.

So values of "cor(e(:,1),e(:,2)) in the order of 1e-3" are well within
the range of values to be expected. Anything smaller than say 1e-4
would in fact be rather unlikely (P approx = 0.025).

Best wishes,
Ted.

--------------------------------------------------------------------
E-Mail: (Ted Harding) <address@hidden>
Date: 15-Feb-99                                       Time: 13:14:19
------------------------------ XFMail ------------------------------