[Rd] Kinderman-Ramage (PR#2846)

Martin Maechler maechler at stat.math.ethz.ch
Sat Apr 26 18:53:06 MEST 2003


>>>>> "PD" == Peter Dalgaard BSA <p.dalgaard at biostat.ku.dk>
>>>>>     on 25 Apr 2003 14:37:58 +0200 writes:

    PD> Prof Brian Ripley <ripley at stats.ox.ac.uk> writes:
    >> Unfortunately it was the default prior to 1.7.0, so for
    >> reproducibility I think we can't change this.  We could
    >> add a new option, a corrected version, but is it worth
    >> it?  I would definitely add a warning if the buggy
    >> version was selected, and perhaps make it an error to
    >> select it via RNGkind (rather than RNGversion or a saved
    >> .Random.seed).

    PD> Argh...

    PD> We do normally correct errors leading to incorrect
    PD> values. I mean, if we found a bug in qnorm() we wouldn't
    PD> hesitate to fix it even if though it affects the
    PD> Inversion RNG, would we?

    PD> How about this: Keep the old generator, but call it
    PD> "K-R-bug" or so.  Fix the real K-R. Make K-R-bug the
    PD> version set by RNGversion. One or more of the methods to
    PD> set it will generate warnings.

This proposal sounds good to my ears.  Since we haven't heard
other comments -- this should go into R-patched, right?
{and who from R-core does it?}

Martin



More information about the R-devel mailing list