[Rd] RNGkind() state (PR#12567)

[Ray Brownrigg]

Prof Brian Ripley wrote:
> On Tue, 19 Aug 2008, Ray.Brownrigg at mcs.vuw.ac.nz wrote:
> 
>> I sent this to R-devel early last month, but have received no 
>> response, so I guess it
>> really is a bug.
> 
> Or people were busy: I was away teaching a course, for example.
> 
>> This looks like a bug to me, and is a bit hard to describe, but easy 
>> to reproduce.  
>> Basically, if RNGkind is saved as something other than the default, 
>> and if the first
>> operation in a session is a set.seed(), the default is reverted to. 
>>  Reproduce by:
> 
> A better description would certainly have helped!  You can look at 
> .Random.seed to see what is happening.
> 
> I think what you are observing is that if .Random.seed has not used in 
> the session, any restored value is not used to set the default kind in 
> set.seed.  I don't think that is unreasonable (in particular as you can 
> set the kind in set.seed() and I take the notes on ?RNG to suggest 
> that), but would be easy to change/document.
> 
My interpretation of what looking at .Random.seed shows is that set.seed() ignores the RNG 
{\bf state} that is contained in the restored .Random.seed (until after the first random 
number has been generated).  ?RNG starts with ".Random.seed is an integer vector, 
containing the random number generator (RNG) state for random number generation in R. It 
can be saved and restored, ...".

I feel it is a bit of a stretch to claim that "If it is NULL, *return* the currently used 
RNG." suggests "if it is NULL, and .Random.seed has not been used in the session, ignore 
the information contained in .Random.seed and *set* the default RNG".

However, knowing what does happen, I acknowledge that the documentation of set.seed() is 
not inconsistent, but I can't accept that the description of .Random.seed would lead one 
to anticipate what does happen.

Regards,
Ray Brownrigg
>>
>> cafe-rozo>  R --vanilla
>>
>> R version 2.7.1 (2008-06-23)
>> Copyright (C) 2008 The R Foundation for Statistical Computing
>> ISBN 3-900051-07-0
>>
>> R is free software and comes with ABSOLUTELY NO WARRANTY.
>> You are welcome to redistribute it under certain conditions.
>> Type 'license()' or 'licence()' for distribution details.
>>
>> R is a collaborative project with many contributors.
>> Type 'contributors()' for more information and
>> 'citation()' on how to cite R or R packages in publications.
>>
>> Type 'demo()' for some demos, 'help()' for on-line help, or
>> 'help.start()' for an HTML browser interface to help.
>> Type 'q()' to quit R.
>>
>>> sessionInfo()
>> R version 2.7.1 (2008-06-23)
>> i386--netbsdelf
>>
>> locale:
>> C
>>
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>> set.seed(1); runif(1)         # check the default
>> [1] 0.2655087
>>> RNGkind("Mar")
>>> RNGkind()
>> [1] "Marsaglia-Multicarry" "Inversion"
>>> set.seed(1); runif(1)         # now check again
>> [1] 0.006153224
>>> set.seed(1)                   # save seed and RNGkind
>>> q("yes")
>> cafe-rozo>  R
>>
>> R version 2.7.1 (2008-06-23)
>> : etc
>> [Previously saved workspace restored]
>>
>>> runif(1)              # with a saved seed, Marsaglia-Multicarry is used
>> [1] 0.006153224
>>> set.seed(1); runif(1)
>> [1] 0.006153224
>>> q()
>> Save workspace image? [y/n/c]: n
>> cafe-rozo>
>>
>> R version 2.7.1 (2008-06-23)
>> : etc
>> [Previously saved workspace restored]
>>
>>> set.seed(1); runif(1)         # saved RNGkind is ignored ...
>> [1] 0.2655087
>>> RNGkind()             # and overridden ...
>> [1] "Mersenne-Twister" "Inversion"
>>> q("no")
>> cafe-rozo>  R
>>
>> R version 2.7.1 (2008-06-23)
>> : etc
>> [Previously saved workspace restored]
>>
>>> RNGkind()             # and wrongly reported!
>> [1] "Mersenne-Twister" "Inversion"
>>> runif(1)                      # Note this is actually using 
>>> Marsaglia-Multicarry
>> [1] 0.006153224
>>>
>>
>> Ray Brownrigg
>> Mathematics, Statistics and Computer Science
>> Victoria University of Wellington
>> Wellington, NZ
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>