Skip to content

pgeom accuracy (PR#6792)

5 messages · terra@diku.dk, Peter Dalgaard, Martin Maechler

#
Full_Name: Morten Welinder
Version: snapshot
OS: 
Submission from: (NULL) (65.213.85.218)


This should fix the remaining two 1-p cancellation issues.

  double l_rt = log1p (-p) * (x + 1);
  if (log_p)
    return R_DT_Clog (l_rt);
  else
    return lower_tail ? -expm1 (l_rt) : exp (l_rt);
#
terra@gnome.org writes:
Never mind the actual content, but snapshot of what? You are being
lazy in a way that potentially causes more work for other people.  

Please state the version exactly, including the date (of the snapshot,
not the submission!) and whether it is of R-devel or R-patched.
#
Apologies.  Snapshot from

    rsync --stats -z -rC --delete rsync.r-project.org::r-devel R

yesterday which contains the fragment pow(1-p,...) which is problematic
for small p.

(If the snapshot I get with the above command has an official name I'll
be happy to use that in future reports.)

Morten
#
terra@gnome.org (Morten Welinder) writes:
The sources should be time-stamped. Take a look at

?version
?bug.report
#
PD> terra@gnome.org (Morten Welinder) writes:
    >> Apologies.  Snapshot from
    >> 
    >> rsync --stats -z -rC --delete
    >> rsync.r-project.org::r-devel R
    >> 
    >> yesterday which contains the fragment pow(1-p,...) which
    >> is problematic for small p.
    >> 
    >> (If the snapshot I get with the above command has an
    >> official name I'll be happy to use that in future
    >> reports.)
    >> 
    >> Morten

    PD> The sources should be time-stamped. Take a look at

    PD>  ?bug.report
    PD>  ?version

if Morten has compiled the snapshot version,...

Otherwise, use  
	   <source>/VERSION
and	   <source>/date-stamp