On 29/12/07 15:38, "Andrew Kappy" <[EMAIL PROTECTED]> wrote:

> I'm looking for help with two random number problems.
> 
> 1. I need to generate 750,000 alpha-numerics, 6 characters long,
> excluding certain letters. I have no problem generating as many random
> numbers as I want, using the RANDOM function, but getting down to 6
> alpha-numerics without duplicates is the problem. I've generated 1.5
> million random numbers and translated them to integers. I've also
> built a numbered table of included alpha characters. I've tried
> various calculations to grab 4 numbers from the larger random number
> and combine them with two of the letters via calculations. The result
> is way too many duplicates.
> 
> 2. I need to generate over 10,000 records of 24 fields, with the
> numbers 1-24 (or letters A-X) randomly distributed through each
> record, with no repeats within a record and no distribution repeated
> within the 10,000 records.
> 
> Any suggestions will be greatly appreciated.

Andy

I think you've had plenty of suggestions as to how this might be achieved.
However, I just wanted to draw attention to one issue -- probably an
insignificant one.

You suggest that you want a set of random numbers without duplicates. If you
REALLY want random numbers, then you cannot apply such conditions. A set of
random numbers may include duplicates; you cannot specify that there are
"too many" duplicates in a particular set of random numbers.

Now, for your purposes, you are quite likely happy with a looser meaning of
the word random; that's fine. I just wanted to point out that you are not
going to end up with a true set of random numbers -- the distribution will
have a very, very slight bias. On the other hand, I believe it's also true
to say that you won't end up with true random numbers if you use the RANDOM
function. I'd suppose the only way to be sure of a really random number is
to base it on something like radioactive decay.

Steve

Reply via email to