[Rd] reproducing Box-Muller numbers

Prof Brian D Ripley ripley@stats.ox.ac.uk
Tue, 22 Feb 2000 21:56:18 +0000 (GMT)


On Tue, 22 Feb 2000, Paul Gilbert wrote:

> Brian
> 
> >I'll make a call to set.seed clear the saved value. That's all we intend to
> >be reproducible, setting the state by set.seed.
> 
> I'm  not sure what you mean by this. Are you suggesting that setting
> .Random.seed directly is not intended to give reproducible results? I presume
> not, but if so, then set.seed needs better control over the seed (i.e. vector
> seeds).

Users are not expected to set .Random.seed directly, period. set.seed()
sets one of 2^32 separate seeds: why do you need more?

> >> Also, could "user-supplied" be added as an option for normal.kind in RNGkind.
> 
> >> I'm sure the Box-Muller in R is better than my own attempt, but I would like
> to
> >> be able to reproduce results I've obtained in the past.
> 
> >Not very easily, and we are in feature freeze for 1.0.0 now, so not before
> >1.1.0. It would be easy to patch in your own version, but the
> >difficulty is to add all the protection needed for user-supplied code.
> 
> When are you expecting 1.1.0? I can easily patch the version used by my library,
> but for obvious reasons I would prefer not to post a library that does this. A
> very large number of my code tests do not work if I cannot reproduce previous
> rnorm random numbers.

1.1.0 is not yet announced, but the feature freeze for 1.0.0 has been for
several weeks (on developer.r-project.org). I presume the numbers you want
to reproduce were not standard R before?

-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-devel mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-devel-request@stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._