[R] Random number generation

plummer at iarc.fr plummer at iarc.fr
Thu Jun 1 22:02:38 CEST 2006


That certainly sounds like the behaviour you would get if you
had a .Random.seed in your work space.  If you do not save your
workspace at the end of the session then the random seed will
be in exactly the same state every time you start a new R session,
and you will get identical simulations from the same sequence
of function calls. So you either need to save your work space
or launch R from the command line with --no-restore or --vanilla
and get a fresh starting seed in each session based on the
time stamp.

Have you tried checking for the existence of .Random.seed before
you call any random number functions (e.g. with
objects(all.names=TRUE))?

I don't know why you would get different behaviour from R 1.8.0
since the RNG hasn't changed since 1.7.0.

Martyn

Quoting Carl <carl at mcs.st-and.ac.uk>:
> Hi All.
> (This is probably an incredibly stupid question :))
>
> I have just noticed that every time I start an R2.3 session and request
> a random number (say with rnorm) that I receive the same number. AFAIK I
> have not set a seed at any point. Returning to an older version of
> R(v1.8) I get the behaviour that I have come to expect, where a new seed
> is automatically generated for each new session (actually it was a
> student running bootstraps on multiple machines that has noticed this).
>
> I am running R2.3 for windows on winXP pro, AMD Athlon 2400+.
>
> Thanks in advance and be gentle.
> C
>
> <...retreats and waits for imminent admonishment...>
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> Dr. Carl Donovan
> Research Fellow in Statistical Computing
> Ph +44 1334 461802
> The Observatory
> Buchanan Gardens
> University of St Andrews
> St Andrews
> Fife
> KY16 9LZ
> Scotland
>
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
>



-----------------------------------------------------------------------
This message and its attachments are strictly confidential. ...{{dropped}}



More information about the R-help mailing list