R-alpha: compatibility

Luke Tierney luke@stat.umn.edu
Mon, 15 Sep 1997 11:55:02 -0500 (CDT)

This is one point where compatibility is not necessarily a good
thing. S, like most systems gives, you no real control over the random
number generator expect re-seeding. Usually this is probably good. But
most simulation books recommend running any simulations you care about
with two very different generators. Most people don't do this because
one generator is usually easy to use (the built-in one in your
favorite system) and anything else is impossible. Another issue that S
faces has to do with running separate bits of a simulation on separate
workstations. Unless you are careful, the separate bits will share
parts of random number streams, and being careful is not really
possible with S's generator (I think) -- it is possible but not easy
with R's. (Of course overlap can be a good thing if you know you have
it -- the point here is that lots of folks run separate simulations
thinking they have independent runs when they probably do not in any
reasonable sense).

The bottom line is that it would be nice to eventually provide "power
users" with more control over the random number generator system.  Now
is clearly not the time to think about how to do that, but it would be
unfortunate if decisions were made at this point in the name of
compatibility that might limit the freedom of doing something "right"
later on.

Just my 2c.


Luke Tierney
University of Minnesota                      Phone:           612-625-7843
School of Statistics                         Fax:             612-624-8868
206 Church Street                            email:      luke@stat.umn.edu
Minneapolis, MN 55455 USA                    WWW:  http://www.stat.umn.edu
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