Re: random number generator

[John Darrington]

On Thu, Mar 26, 2009 at 08:52:33AM +0100, Matej Kovacic wrote:
     
     > It works for me.  Perhaps you forgot to hit the "Run Pending
     > Transformations" button?
     
     Exactly that was the problem. Well, this is something which always make
     me angry also in in new SPSS versions. :-))

In defence of this design, it means that you can avoid unnecessary
passes through the data, which on large datasets can severely affect
performance. 

     
     BTW, I have found that when you are creating samples in SPSS, something
     strange happens - if you create let's say 5 samples of the same size,
     SPSS is selecting the same units (I found up to 20% of the same units in
     this 5 samples).
     
     It seems SPSS is unsing faulty (biased) random number generator,
     probably RNG from Windows (I was using Windows 2000, which is known to
     have weak RNG).
     
     What exactly is using PSPP?

We use the Mersenne Twister implementation from the GNU Scientific
Library.   I believe later spss versions allow you to choose either
the Mersenne Twister or the buggy windows RNG.

J'

-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

signature.asc
Description: Digital signature