Re: random numbers in fortran

On 2006-11-30 11:03:14 -0400, Herman D. Knoble <SkipKnobleLESS@xxxxxxxxxxxxxxx> said:

Gordon: I agree with you about seeking information from statistics and mathematical news groups. I disagree with that the value of 52! has any PRACTICAL bearing on this simulation at hand; that is, it is not necessary to COMPUTE 52! for simulations involving a card shuffleing
once a "reputable" RNG is available (which they are as outlined
in one of my postings). I also disgree that comp.lang.fortran
is not the best place to seek such advice.

The point is to illustrate the notion that card shuffling is not as trivial
as many suppose it is. There are 52! possible permutations of 52 objects.
If the simulation is to be scrupulously honest then it should be capable of
producing all of these in some order. So a cycle length that provides all
the possible ordered blocks of size 52 would seem to be in order. Cycle length
is not the end all but it at least says the technical considerations are
present.

Now that is a tall order. But unless the origninal poster understands that
there is an issue he will only be providing "games for young children". The
professional judgement comes in figuring out where the tallness of the order
can be cut short. The original question was shuffling and the implication
was for results on card play. The story has since been revised.

(In this case there are generators that have been tested to be equidistributed
in 52 (or more) dimensions which one would expect to be adequate. But the random
intrinsic can not be expected to be one.)

When the poster is asking how to cheat short cycle length by using various
seeds it seems that there is a lack of professional judgement.

Asking about random number implementations in c.l.f is fair game. Asking
about the statistical theory behind simulations is hoping for very good luck.
It is often provided for random numbers, numerical linear algebra, diferential
equations solving and a variety of other topics but it is still good luck
and the tolerance of the c.l.f readers. In practice it is often a good
place to get pointers. Treating c.l.f as a definitive source rather than a
convenient and quick source of practical advice that may lead to definitive
advice is to misunderstand things.



.

Relevant Pages

  • Re: random numbers in fortran
    ... I agree with you about seeking information from statistics ... ANY Monte Carlo simulation is better done with a statistically better ... quasi-random number generator. ... -|> Generally in practice then, could I not compensate for a short cycle by ...
    (comp.lang.fortran)
  • Re: My wife getting scared
    ... Part of a successful simulation is in the mind of the pilot. ... and then go out and fly the airplane to test standards. ... For some emergencies, the best possible response is an automatic, by-rote ... How often do you practice these things in full-motion simulators? ...
    (rec.aviation.piloting)
  • Re: No Pseudo-Pilots
    ... There was no advance practice in real spacecraft prior to actually doing it. ... Fortunately, simulation does the job. ... That may even provide a nebulous foundation for an opinion. ... derived on that knowledge base lacks any real credibility. ...
    (rec.aviation.student)
  • Re: recommendations for exam simulations (failed 292)
    ... to come together and form the follwing words in: ... nothing but practice speed in the 'real' test environment. ... get another Microsoft approved test simulation. ...
    (microsoft.public.cert.exam.mcsa)
  • c-mex dynamic structure size
    ... I came upon a problem I cannot solve on my own and therefore I'd like to as this group's users for an advice. ... I use Matlab C-mex files to communicate with a water system simulation package. ... I've tried allocating a new structure in each iteration step and copying the data from the previous structure into the new bigger one. ...
    (comp.soft-sys.matlab)