This is a very simple change at C level: the !R_FINITE(p) test is
incorrect for log_p != 0.
Changed now.
On Wed, 23 Nov 2005, Spencer Graves wrote:
> Hi, All:
>
> For most but not all cases, qbinom is the inverse of pbinom.
> Consider the following example, which generates an exception:
>
> > (pb01 <- pbinom(0:1, 1, .5, log=T, lower.tail=FALSE))
> [1] -0.6931472 -Inf
>
> Since "lower.tail=FALSE", Pr{X>1} = 0 in this context, and
log(0) > -Inf, consistent with the documentation.
>
> However, the inverse of this does NOT recover 0:1:
>
> > qbinom(pb01,1, .5, log=T, lower.tail=F)
> [1] 0 NaN
>
> Shouldn't the NaN here be 1? If yes, this is relatively easy to
fix.
> Consider for example the following:
>
> qbinom. <-
> function (p, size, prob, lower.tail = TRUE, log.p = FALSE){
> q. <- .Internal(qbinom(p, size, prob, lower.tail, log.p))
> q.[p==(-Inf)] <- 1
> q.
> }
> > qbinom.(pb01,1, .5, log=T, lower.tail=F)
> [1] 0 1
> Warning message:
> NaNs produced in: qbinom(p, size, prob, lower.tail, log.p)
>
> It's also easy to eliminate the Warning. Consider for example the
> following:
>
> qbinom. <-
> function (p, size, prob, lower.tail = TRUE, log.p = FALSE){
> if(any(p.inf <- p==(-Inf))&&(!lower.tail)&&log.p){
> n <- max(length(p), length(size), length(prob))
> p <- rep(p, length=n)
> size <- rep(size, length=n)
> prob <- rep(prob, length=n)
> q. <- size
> q.[p>(-Inf)] <- .Internal(qbinom(p[!p.inf],
> size[!p.inf], prob[!p.inf], lower.tail, log.p))
> return(q.)
> }
> .Internal(qbinom(p, size, prob, lower.tail, log.p))
> }
>
> I suspect that for the right person, it would likely be easy to fix
> this in the .Internal qbinom code. However, that's beyond my current R
> skill level.
>
> Thanks for all your efforts to make R what it is today.
> Best Wishes,
> spencer graves
>
> --
> Spencer Graves, PhD
> Senior Development Engineer
> PDF Solutions, Inc.
> 333 West San Carlos Street Suite 700
> San Jose, CA 95110, USA
>
> spencer.graves at pdf.com
> www.pdf.com <http://www.pdf.com>
> Tel: 408-938-4420
> Fax: 408-280-7915
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>
--
Brian D. Ripley, ripley at 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 272866 (PA)
Oxford OX1 3TG, UK Fax: +44 1865 272595