Hi Bernd,

Suggest to dovetail discussions into the OP's thread in
..worksheet.functions,
http://tinyurl.com/ab6t2z

I've just posted my last follow-through there with a way to address the OP's
non-dups requirement.

You might wish to re-post your comments/alternative offers to the OP over
there.

.. Your example does not ensure this ..
Agreed, and I did mention this explicitly in my response detailing the
modeling steps to the OP.
--
Max
Singapore
http://savefile.com/projects/236895
Downloads:23,000 Files:370 Subscribers:66
xdemechanik
---
"Bernd P" wrote in message
...
Hello Max,

1. I think the OP asks for exactly the 4/12 of ABC hits in the output
not only 4/12 likehood. So you would need something like the D'Hondt
election distribution approach to define the number of ABC hits in the
output (easy for 50% ABC and 50% XYZ on a selection of 100 but what if
you have 99? You can only take 49 ABC's and 49 XYZ's and randomly
distribute the remaining one, I guess). The initial unique list and
frequencies you can get with my UDF lfreq, for example.

2. The OP obviously does not want to come up with dupe ref# at the
end. Your example does not ensure this. I suggest to take my UDF
UniqRandInt for this.

Regards,
Bernd

www.sulprobil.com