[Rd] Bias in R's random integers?

[Carl Boettiger]

Dear list,

It looks to me that R samples random integers using an intuitive but biased
algorithm by going from a random number on [0,1) from the PRNG to a random
integer, e.g.
https://github.com/wch/r-source/blob/tags/R-3-5-1/src/main/RNG.c#L808

Many other languages use various rejection sampling approaches which
provide an unbiased method for sampling, such as in Go, python, and others
described here:  https://arxiv.org/abs/1805.10941 (I believe the biased
algorithm currently used in R is also described there).  I'm not an expert
in this area, but does it make sense for the R to adopt one of the unbiased
random sample algorithms outlined there and used in other languages?  Would
a patch providing such an algorithm be welcome? What concerns would need to
be addressed first?

I believe this issue was also raised by Killie & Philip in
http://r.789695.n4.nabble.com/Bug-in-sample-td4729483.html, and more
recently in
https://www.stat.berkeley.edu/~stark/Preprints/r-random-issues.pdf,
pointing to the python implementation for comparison:
https://github.com/statlab/cryptorandom/blob/master/cryptorandom/cryptorandom.py#L265

Thanks!

Carl
-- 

http://carlboettiger.info

	[[alternative HTML version deleted]]